Thursday, October 6, 2011

Tool: Google Cloud SQL: your database in the cloud

Free mysql in the cloud

Google Cloud SQL: your database in the cloud

Author Photo

By Navneet Joneja, Product Manager for Google Cloud SQL

Reaction: Over Two Decades of Integration Based, Geometric Flow Visualization

The paper attempts to classify the geometric flow-visualization approaches. The author discusses the challenges in  the field of flow visualization and the classification itself is based on the challenges. The idea of explaining terminologies like  streamline, streakline, paths and surfaces before hand has made the understanding easy and clear. The example of tornado visualization is very interesting. It depicts us how rich a dataset can be and how different visualization techniques can help us perceive multiple dimensions of it.

Overall, this paper has been a great read to get introduction of flow-visualization. Towards the end, the author states that "as the dimensionality of the integral object increases the volume of research decreases" which makes me wonder whether some of the challenges mentioned by the author may not be actually due to lack of computing power.

Wednesday, October 5, 2011

Reaction: Marching cubes: A high resolution 3D surface construction algorithm

     This paper talks about marching cubes  Algorithm " A new, high-resolution 3D surface construction algorithm. Basically the algorithm has following steps

1.Read four slices
2. Scan two slices and create a cube
3. Calculate an index
4. Find the list of edges from LUT
5. Find the surface-edge intersection
6. Calculate a unit normal at each cube vertex
7. Output: the triangle vertices, vertex normals." The adaptive algorithm produces a very close approximation of the original surface.New methods that extend marching cubes algorithm will likely produce more accurate localization.

Reaction : Over Two Decades of Integration-Based, Geometric Flow Visualization

    This paper talks about the challenges inflow visualization. "Fluid Visualization in  is used to make the flow patterns visible, in order to get a qualitative or quantitative information on them."It lists out the problems like huge volume of data, Seeding and perception. It discusses about research that took place in this field and explains various techniques. The paper says that none of the methods are perfect. There are challenges like the user's goal, huge temporal data.Different seeding techniques are used commercially, streamline technique is majorly adopted as it is easier.The paper concludes that there are lot of problems like unsteady flow visualization yet to be solved and more work should be done. 

Reaction: A Survey of Algorithms for Volume Visualization

This is an excellent read as an introduction of volume visualization. The paper not only explains and compares the Volume Visualization algorithms, it also discusses other aspects like data and volume characteristics. The paper describes two methods of volume visualization DVR and SF. The paper makes a important point under photorealism section that non plausible objects are difficult to interpret and hence we should avoid using such objects. The paper describes different algorithms in brief. The paper gives a good introduction of volume visualization algorithms.

Reaction: Over Two Decades of Integration Based, Geometric Flow Visualization

This paper mainly talks about the geometry based visualization techniques. It also talks about the side of visualization that deals with motion. Today the field of visualization has advanced to a great extent. Motion and user interaction plays an important role in graphical representation. I think it has been adopted universally. And this is where flow visualization comes into play which has been discussed briefly in this paper. The author also talks about the various challenges faced in flow visualization. I think the biggest challenge now a days is the length and vastness of the data sets. They are so huge that a lot of time is consumed in deciding how to display the available data.

Then the author has discussed various techniques in flow visualization like the direct method, texture based LIC techniques and geometric method which is based on the geometry of objects and last but not the least the feature method.

Another interesting discussion according to me was the example of the tornado using stream surface in comparison to flow volume.

RIP Steve Jobs

It's a sad day in the history of technology...

Reaction: Imaging Vector Fields using Line Integral Convolution

This paper talks about the linear and curvilinear filtering techniques to locally blur textures along a vector field. There is a good comparison made in this paper between DDA convolution and Line integral Convolution. The DDA convolution renders the vector fields unevenly , treating linear portions of the field more accurately than small scaled vertices. This is undesirable for some vector fields as the details in small scaled structures are lost. So for this there is an alternate technique used in the paper called the LIC. This technique uses a completely different approach for approximating a vector . In LIC the author has discussed different things such as periodic motion filters, normalization, implementation and application. LIC has several advantages that have been discussed in the paper like removing the aliasing error, a detailed description of the data and the ability to interface different techniques with LIC.

Reaction: A survey of Algorithms for volume visualization

Volume visualization, as the paper suggests, is one of the most fast growing areas in the field of visualization. Many graphics techniques are used in volume visualization. This paper discusses the way in which these techniques are applied in fundamental volume visualization and the advantage and disadvantages of the algorithms used in those techniques.

In this paper different algorithms like SF, DVR, Ray Casting, Marching Cube etc algorithms have been discussed. Of all the algorithms I found ‘splatting’ which comes under DVR the most interesting one. Splatting performs a front to back object-order traversal of the voxels in the volumetric dataset. Each voxel’s contribution to the image is calculated and composited using a series of table lookups. It is called splatting because it is similar to throwing a snowball on a glass plate.

The section on photorealism is also very interesting. Here they have questioned if it is appropriate to visualize non plausible objects as people would find it difficult to relate to it. Well I agree to some extent but on the other hand I have never seen a visualization representing a physically non plausible object so I don’t really know how difficult an interpretation would be. Overall it was a nicely written paper with some interesting algorithms.

Reaction: Marching Cubes: A high resolution 3D surface construction algorithm

This is a very interesting paper that talks about the 3D surface construction algorithm. Basically this model helps to produce an unprecedented detailed models from a 3D array data.

Then the author talk about the information flow for 3D medical algorithms which includes the steps like information acquisition, image processing, model creations, viewing operations and displaying the surface using techniques like ray casting, depth shading and color shading.

Then comes the actual marching cube algorithm which has been explained with the help of several diagrams. There is also a mathematical representation of this algorithm which was a bit confusing but the explanation was enough for me to understand the actual algorithm. Being a student of computer graphics I could easily relate myself to this particular paper as I have read about this particular algorithm earlier as well.

Reaction: Imaging Vector Fields Using Line Integral Convolution

The paper focuses on Line Integral Convolution, an algorithm for imaging vector fields. The paper gives concise explanation and mathematical model of the algorithm. It also compares the images created by this algorithm with those created by DDA convolution, to depict the effectiveness of the algorithm. The paper explains how integration of periodic motion filters with LIC can produce the motion effect. It also discusses how images can be fine tuned using normalization technique along with LIC. Some of the example images like checkerboard and Flavian Amphitheater under post-processing section are particularly interesting.

Reaction: Marching Cubes: A high resolution 3D surface construction algorithm

This paper discusses a 3D medical data processing algorithm, The Marching Cubes. This algorithm divides the visualization surface into unit cubes and marches through them using a divide-and-conquer strategy. For each cube, the algorithm determines how surface to be visualized cuts through the cube and then moves to the next cube. Most of this explanation in the paper comes through complex mathematical formulae and terms. The author also gives example visualizations drawn using the algorithm. However, the effectiveness of the algorithm could have been underlined if the results of marching cube were compared with those from other similar algorithms.

Reaction: Over Two Decades of Integration-Based, Geometric Flow Visualization

The paper talks about flow visualization which is interesting because it helps identify flow features and characteristics. I liked the way in which the paper spoke of the challenges faced by flow visualization. The paper introduces several interesting terminologies such as streamline and streakline. Streamlines used to visualize 3D vector fields forms an informative read.

Reaction: Over two decades of Integration-based, Geometric Flow Visualization

This paper introduces the challenges faced today in terms of computing power required to process multi-variant geometric flow visualizations. These flow visualizations have been surveyed and classified by the authors according to the challenges, which helped me a great deal in understanding these visualizations. This paper is a good starting point for learning the basics of spatial visualizations and getting some ideas about the algorithms used in flow viz. research. Even though there have advancements in the computing power, I was surprised to read the authors state that there are still unsolved problems, after 20-25 yrs of research.
There are some interesting terms introduced by the authors, such as streamtube and stream surface. The most interesting discussion that caught my eye was on how different the simulation of a tornado using stream surface was in comparision with flow volume, even though the dataset was the same.

Reaction: Imaging vector fields using line integral convolution

A vector field can be thought of as a vector to each point in a plane. This understanding of the vector field was not apparent while reading the paper. The paper talks about mapping these vector fields into two and three dimentional space using Line Integral Convolution. One thing to notice is that the paper talks about deriving the method of line integral convolution from DDA convolution which is informative. The pictures depicting the variation of Local streamline L are very informative.

Reaction: Marching Cubes: A high resolution 3D surface construction algorithm

Marching cubes is an algorithm that processes 3 dimensional medical data and builds models with fine details. By specifying density values, the users can select the desired surface. This algorithm basically works on the technique of divide and conquer and is greatly useful in the healthcare software industry. This is a somewhat old paper and it would be interesting to know about the latest developments in the models built using this algorithm. On the positive side, it was interesting to learn about its application to medical softwares on account of the diagrams presented by the author such as bone surface, soft tissue, etc, although these figures are not clearly presented in the paper. One suggestion that came to my mind regarding the paper was to include a comparision with the different models in existence at the time this paper was published, just to give an idea of the effectiveness of this algorithm.

Reaction: Imaging Vector Fields using Line Integral Convolution

This paper introduces a new algorithm to use curvilinear filtering techniques to locally blur textures along a vector field. Algorithms which image directional information can have wide applications. This technique can also produce some interesting novel effects.
This paper gives a nice introduction to the domain of directional information. I feel that motion filters are fast catching up due to hardware advancements and this is going to open up new vistas in visualization in the future. It remains to be seen when this actually takes place and how the user experience in visualization will be enhanced at that time.

Reaction: A Survey of Algorithms for Volume Visualization

In this paper, the two main categories of volume visualization algorithms - SF and DVR have been presented, alongwith the different algorithms such as Ray Casting, Marching Cubes, etc. This spans the range from cube techniques for surface fitting to projection and image-order methods.
The discussion on photorealism is aptly presented and is a good value addition. I found the argument very interesting. Firstly, it states that visualization should be bounded by common interpretation. Secondly, it says info viz. is about abstracting data to provide that amount of dataset which can be easily interpreted. On the other hand, some concepts were a bit difficult to understand, such as the depth fog and brightness attenuation. Nevertheless, the paper was well written and helped me understand a great deal about the different techniques and challenges that usually define volume visualization.

Reaction: Over Two Decades of Integration-Based, Geometric Flow Visualization

The paper discuss about flow visualization. Author discuss several challenges encountered like large datasets, interaction,perception etc and major advancements made to overcome these challenges over decades. As a result of which flow visualization now has its application in many areas.Diffrences between interactive and automatic seeding and techniques to perform geometric visualization has been expalined well. Flow Visualization has been contributing significantly in the field of medical, mechanical and aviation, and thus hold scope of improvements. Overall its an interesting read.

Reaction: Imaging Vector Fields using Line Integral Convolution

The paper discuss about two main onvolution techniques DDA and LIC,diffrences in approachs, their performance considerations and usage.DDA approximate local vector using straight line whereas LIC relies on a central location from where rendering can be done in diffrent direction.Both the alogrithms have applications in diffrent areas like motion imaging and rendering linear fields. LIC is definately provides better quality but at the cost of performance and speed as LIC is believed to be slower than DDA. Paper is well-organized and supported with examples and images but clear understanding. It also leaves scope for lot of improvements.

Reaction: A Survey of Algorithms for Volume Visualization

This paper discuss about importance of Volume visualization and fundamental alogrithms like contour-connecting, Opaue cube, Marching cubes etc which explains volume visualization at implementation level.Contents are well organized and systematic and thus helped me in understanding the concept of volume rendering step-by-step.Concepts like Photorealism or Marching tetrahedral or Contour connecting should have been explained more with good practical example, as these concepts require understanding.overall its an interesting read.

Reaction: A Survey of Algorithms for Volume Visualization

This paper provide lot of knowledge about different volume rendering techniques along with their trade offs. It was good to read the so-called foundation algorithms in the field of volume visualization. The paper is very well compiled and organized and this make it easier to grasp the information given in it. Some of the techniques like feature extraction, nullify certain pixels, noise cancellation were fairly easy to understand as I have already used them in one of my projects called 'Iris Recognition system'. I liked and totally agree with the portion of the paper which talks about the importance of data collection and classification when it comes to volume rendering. Also having read marching cubes in one of the other papers made it quick to go through this chunk of the paper. Overall I found this paper to be very useful and informative.

Reaction: Imaging Vector Fields Using Line Integral Convolution

This paper focuses on an effective new approach that is to use linear and curvilinear filtering techniques to locally blur textures along a vector field which is capable of imaging arbitrary two and three dimensional vector fields. The technique can also produce novel special effects. Line integral convolution represents a new and general method for imaging two- and three-dimensional vector fields. The algorithm filters an input image along local stream lines defined by an input vector field and generates an output image.

One thing I didn't like about the paper was that its not an easy read as the other papers. Though the mathematical formulas were a requirement, they could have been presented in a much easier format making it a fun to read.

Reaction: Marching Cubes - A High Resolution 3D Surface Construction Algorithm

This paper discuss about a new 3D surface contruction algorithm called marching cubes, significant in examining 3D medical data or anatomy. Author justify the need of this alogrithm by explaining the diffrence in interepretation mechanism of a physician and radiologist.Data acquired in form of 2D slices from techniqus like CT,MR and SPECT, serve as an input to this algorithm.Reason behind failure of early 3d surface construction techniques(eg inter-slice connectivity,cuberilee, ray casting, volume models) was their inablity of holding back useful information in original data. Keeping that in mind, marching alogrithm carefully derives inter-splice connectivity, surface location and gradient vector from original 3D data. Algorithm is primarily divided into two parts:-1) locating surface cordinates 2) creating triangles.It uses combination of techniques like divide and conquer, linear interpolation, indexing and Permutation to locate and calculate triangle vertices and vertex normals.Coherence,which promotes pre-calculation and boolean operations that enable solid modelling are definately significant advancements over existing algorithm. Author explains the practical application of this alogrithm wit the help os case studies on data gathered using CT, MA and SPECT.i believe dividing cubes would be equally intutive. Overall its an interesting read.

Reaction: Marching Cubes - A High Resolution 3D Surface Construction Algorithm

This was an old one given to us. But despite of being an old paper it was really interesting. This paper talks about a 3-dimensional surface construction algorithm called marching cubes, which could produce models with fine details. Actually marching cube uses the technique of divide and conquer, which most of us have read in the design and analysis of algorithms classes. Also having been worked in the health-care software industry made my life bit easier in understanding and correlating some of the work flows. It was good to read about the importance of image processing and visualization in the medical field. Also the good amount of diagrams and pictures used to explain the marching cube algorithm is really helpful in understanding the algorithm as such. Although I don't have much background of computer graphics but still it was fairly easy to go through this one. Overall I was happy to read such an old paper which was published around the time I was just born. :P

Reaction: A Survey of Algorithms for Volume Visualization

Authors have presented the survey by giving an introduction of the fast growing field of volume visualization for the graphics programmer. They have not discussed the graphics techniques, information of which we can get from anywhere but what is explained is the way these techniques are applied in fundamental volume visualization algorithms. Not only the advantages and disadvantages of each algorithm is given but the algorithm's space and time requirements, ease of implementation, and the type of data it handles best is also listed in the peer which is really informative and concise. The survey also reviews the terms, procedures, and heuristics most often used in the field of volume visualization.

I feel they have done quite a good job in putting up all the information together. Even after specifying all possible information, since the field is so large, they have recommended that one should read from the papers listed in bibliography for further information. They have even mentioned which topics they have not covered so as not to confuse reader about anything.

Reaction : Over Two Decades of Integration-Based, Geometric Flow Visualization

In this paper, geometry-based flow visualization techniques form the focus of discussion. The author makes a very good observation that although technology is moving forward there is a gap between data sets and visualization. A key observation is the fact that although there are a lot visualization tools, there is no once such tool that gives good results to the various flowing visualization. Also there is considerable focus on 2D and 3D unsteady flows can be a problem to focus in future.

Reaction: Over Two Decades of Integration-Based, Geometric Flow Visualization

This paper provides comprehensive information about the various approaches and trends over a period of last two decades but it also focuses on a new branch of visualization which deals with motion. It also covers a complete set of problems, solved or unsolved, in the field of geometric flow research. When it comes to motion we want to take into consideration the direction along with the magnitude. This paper introduces to us this new visualization called Flow Visualization. What I liked best about the paper is the fact that the authors have put in efforts to make this paper simple and easy to understand. The tabular overview of researches, the comparison between various methods, their advantages and disadvantages are some of the explicit features I liked about the paper.
To sum up it was a good paper to read because it took me to a new dimension of visualization and was different from the ones we have read so far.

Over Two Decades of IntegrationBased, Geometric Vector Field Visualization

The paper talks about flow visualization, a technique that stresses on the importance of vector data representation in images. There are many applications that require flow visualization like medical visualization, oceanography and simulation.

The author talks about the many challenges that one can face in the process of flow visualization. The challenges include extremely humongous data sets, seeding and placement of objects, computation time for processing. All these factors and more are to be considered before arriving at a visualization technique. The authors enlist and make a comparison about the various techniques present for flow visualization, namely, direct method- which is simple but dos not present high quality images, texture-based-like the LIC technique discussed in one of the previous papers; geometric method that takes the geometry of the objects into consideration and the feature method. The paper discusses how each of these techniques are good in some cases and also mentions the cases to which the technique may not provide the best of results. Overall, this paper acts as a good reference to understand and choose among the various flow visualization techniques.

Reaction: Imaging Vector Fields using Line Integral Convolution

This paper was bit easier to digest for me as it had some stuff similar to what I had already studied in one of my earlier classes. Though the terms like normalization and convolution always irritate me but when read in the context of visualization course it made a lot of sense to me as this paper talks about the image vector processing techniques. The author has thoughtfully given us the necessary beackground for the convolution algorithms. The discussion wherein it talks about the differences, performance considerations and usage of the two algorithms, namely DDA and LIC, seems really important. DDA is shown to be faster than LIC but the better quality of the image given by LIC gives it an edge over DDA. It was also good to learn the LIC could be used in motion imaging for the purpose of debluring. This has left the platform open for some interesting future work. In sum, this paper was a good one to read. Moreover, the clarity with which the concepts have been explained along with before and after images make it a commendable work done by the authors.


In this paper a new, high-resolution 3D surface construction algorithm has been introduced that produces models with much detail. The proposed algorithm, marching cubes, creates a polygonal representation of constant density surfaces from a 3D array of data.

The importance of the algorithm has been very well validated by the authors. They have first described the information flow for 3D medical applications, related work and then the drawbacks of that work. After this they have explained the efficiency and functional enhancements, supported by case studies using three different medical imaging techniques to illustrate the new algorithm's capabilities.

Authors have provided ample number of diagrams to explain marching cubes algorithm. Figure 8. Bone Surface, Figure 9. Soft Tissue Surface, Figure 10. Soft Tissue, Top View and the likewise are not clear enough. Since the paper is old enough it doesn't provide the latest development in the field and hence a little out of date. I am not really sure if reading this paper would benefit anyone other than providing historical information in the field of 3D modelling. But considering that the paper is very old, its amazing that such an advanced algorithm was implemented at that time.

Reaction: Marching cubes: A high resolution 3D surface construction algorithm

The paper presents an algorithm called marching cubes, that processes 3D medical data and build models. The paper talks about users selecting the desired surface by specifying a density values. This seemed a little vague. The paper further describes the marching cube algorithm. The summary was way more understandable than the mathematical representation. It would have been interesting to read about the recent enhancements or techniques to building models. It would also have helped to compare the model with others of its time.

Reaction: Over Two Decades of Integration-Based, Geometric Flow Visualization

The authors of the paper talk about flow visualization, a branch of viz that focuses more on the vector quantities than the scalar, impling that they would want to take even direction into consideration in visualizations apart from the magnitude.

This informative paper does a good job in giving an overview of vector field visualization and its many techniques. It talks about the challenges and parameters to be considered in choosing a suitable technique, which include large data sets, the visualization of both magnitude and direction simultaneously, interactions, computation time and perception. The authors enlist how different visualization techniques can handle these challenges.

What I found fascinating are the different perspectives for the way data is characterized by so many dimensions like spatial, temporal and others like velocity, pressure etc . I also found the geometric flow viz fascinating where discrete objects are computed based on the fact their shape directly relies on the underlying geometry. I wish there were more examples of the cases in which these different techniques could be used though.

Reaction: A Survey of Algorithms for Volume Visualization

The paper mainly talks about visualization algorithms and the many related concepts that are to be considered for volume visualization. I felt this paper is very well written and organized and can act as a brief yet informative read on the techniques and the challenges involved in such high quality, huge visualizations.

Moving on, I could relate and understand this paper well as I have worked on an image processing project that isolates a desired or target object from its background. So , I had to using the technique of feature extraction where a threshold value is applied to choose or nullify certain pixel values. As I read this paper, I found that this is a popular technique and is called the surface-fitting(SF) algorithm.

I also agree with the authors that data classification is one of the greatest challenges in such visualizations. It is an elaborate iterative trial-and-error process to get the right threshold for image properties.

Among the algorithms, we did read marching cubes in detail in one of the papers. However, I found the ray casting algorithm fascinating as I do feel from the description of the method that it might produce the best images. I assume that it is computationally more intensive because of its detailing

Reaction: A Survey of Algorithms for Volume Visualization

The paper talks about the top algorithms upon which volume visualization is based. Volume visualization as a topic is fascinating because it is the basis for building images(3-D) from 2-D data. I found the section about Photorealism an interesting read because it speculates whether visualization should be about only physically plausible objects. I believe both sides have valid arguments. Among the volume visualization algorithms, I found ray-casting an effective technique because no shadows or reflections are generated and the dolphin head picture in the paper depicts its ability to visualize volume.

Reaction: Over Two Decades of Integration-Based, Geometric Flow Visualization

The paper intendes to put forward the field of flow visualization are about viewing discrete objects in a velocity field to characterize flow. Therefore these types of visualization have a temporal component alongwith the spatial component.

The paper gathers all the challenges encountered and the developments made in the last two decades in the field of flow visualization which has great applications in the various pratical fields. The paper classifies approaches into four general categories.

DIrect technologies reply more on vectors data or color according to the magnitude and are simple and efficient to implement. Because of this simplicity, these techniques often require processing before visualization. On the other hand, geometric techniques use trajectories, pathlines to build objects. The article splits the two decades into 2 parts of 10 years each where the former concentrated on particle tracing and the next decade was about seeding technologies.

The first section of the paper gives a detailed classfications of different geometrical methods on which research was carried out in the last two decades alongwith the challenges encountered and solutions found out. This table also highlights that flow visualizations has been a field of great interest and constant developments have been made over the years.

Reaction: Imaging Vector Fields Using Line Integral Convolution

The paper talks about imaging vector field technique that help to improve upon the texture of visualizations and produce high quality images. I did find the paper slightly more complicated than the previous two and mathematical but I agree that it is an interesting read.

The paper explains about how DDA convolution takes the vector field to be a straight line and how this adversely affects the texture and detail of the image in cases where the radius of curvature is small. To overcome this, the LIC (line integral convolution) technique is used to approximate vector fields starting from the center of a pixel.It was fascinating to read how noise is effectively reduced in the whole process to generate lower blur, good quality images. The process of normalization to maintain both brightness and contrast was interesting.It was pretty catchy to read about the implementation of LIC and the impact that the input texture has and how the images are post processed to remove noise. I think I would like to read other related papers to better understand these concepts and the mathematics involved.

Reaction: Marching Cubes - A High Resolution 3D Surface Construction Algorithm

Although there are technological advances in the medical field helping physicians understand the internal bone/tissue structure, it has become essential to ensure that this info is presented in a way that is more visual and easy to comprehend. This paper presents an algorithm called marching cubes that helps the above cause and generates 3D surfaces of input medical data.

The paper is an interesting read and I personally was unaware of the importance of image processing and visualization on medical data. The algorithmic approach presented in the paper is easy to comprehend. The steps of marching cubes algorithm and optimizations at each level are explained very well. The reader will need some background in computer graphics to understand some of the optimizations applied to the algorithm. There are some nice pictures which show the effect of using the algorithm. To do justice to the algorithm there should have been a comparative study with one of the algorithms mentioned in the related work.

Imaging Vector Fields Using Line Integral Convolution

The paper talks about the importance of vector representation in images along with maintaining the texture information. It also talks about how white noise is eliminated or effectively reduced .

The paper compares two different techniques DDA convolution and Line Integral Convolution. DDA convolution assumes a straight line for every vector field point. In this case, the texture information is lost if the objects in the image have a small radius of curvature. In case of such objects, the border appears to be smudged or blurred with the DDA technique. So, the authors talk about an alternative technique called the LIC. The paper talks in detail about how LIC uses a different approach of approximating a vector from the center of the pixel and it adds negative and positive directions to it to preserve texture information. The authors have expressed the steps involved in LIC by applying low pass filters, normalizing the data points to maintain brightness values and post process the image to eliminate extraneous noise. Although LIC is computationally intensive and tedious, by the explanation of the techniques I assume that LIC proves to be a much better technique than DDA, which is a good trade-off to obtain high quality images.

Reaction : Imaging Vector Fields Using Line Integral Convolution

Line integral convolution represents a new and general method
for imaging two- and three-dimensional vector fields. The algorithm
filters an input image along local stream lines defined by an
input vector field and generates an output image. Figure 3 is really an interesting one.A two-dimensional vector field showing the local stream line starting in cell (x, y) shows the direction of the flow. When we scroll we actually can see as thought there is motion. It is mainly because of the arrow in different direction differed by a small degree, when scrolled we see the overlapping of arrows and hence we see a motion in the direction of the arrows.

Reaction: Marching Cubes: A High Resolution 3-d Surface Construction Algorithm

This paper provides good background information by describing the uses of CT, MR, SPECT and how it aids physicians in understanding them. I appreciate the fact that such an algorithm was designed back in the 80's and I am most certain that it has many applications today and maybe even many flavors of the algorithm. The models produced by the marching cubes algorithm have a striking similarity of reality because of the amount of detail that can be visualized in the models.
The calculation of cubes from the input slices seems like an improvement over the older algorithms where the calculations are not fine, but marching cubes is formulated such that every detail of the image is taken into consideration. The improvements the authors give at the end using dividing cubes can a good enhancement to represent big volumes of 3-d data.

Reaction: Imaging Vector Fields Using Line Integral Convolution

The paper starts off with introducing some interesting techniques to image vector fields. It introduces two major techniques DDA and LIC. The former assumes that the local vector can be approximated using a straight line while the latter assumes it starts at the center of a pixel and moves out in positive and negative directions. That is the reason why DDA is able to render linear fields more accurately. Illustrations show that LIC is accurate in displaying textures and vector fields. The paper has supported LIC technique and its output through before and after images.

LIC has the following advantages:
1. Presents data in a more detailed manner.
2. Can removing aliasing efficiently.
3. Different techniques can be interfaced with LIC.
4. LIC can be generalized to higher dimensions.

Results say that LIC is about 10 times slower than DDA. It would be interesting to see how accurate LIC is after techniques to determine their efficiency are invented.

Reaction: Over Two Decades of Integration-Based, Geometric Flow Visualization

Authors have reviewed and classified geometric flow visualization literature in this paper. The overall result is a complete collection of solved and unsolved problems in the rapidly evolving branch of geometric flow research. They have analyzed the latest developments and introduced a novel classification scheme based on challenges including seeding. The tabular overview of research and highlights of unsolved problems as well as challenges along with a range of solutions have been provided and authors have put up tremendous effort in it. They have not simply provided an enumeration of related papers in integration-based, geometric flow visualization, but also compared different methods, related them with one another and weighed their relative pros and cons.

The literature has been organized based upon the dimensionality of the geometry-based objects in the domain and the dimensionality of the data domain itself. This organization of literature points to the mature areas where many solutions are offered and even those areas which have a lot of unsolved problems. I even got to know about lot of new terminologies like streamtube, stream ribbon etc through this paper.

Its a good resource for people interested in knowing more about different algorithms implemented in the field of flow visualization.

Reaction: Imaging Vector Fields Using Line Integral Convolution

The importance of direction in visualizing the data is realized and the application of this type of data in various fields is significantly improved with the use of line integral convolution. the author shows the drawbacks of other forms of vector fields imaging like DDA convolution and spot noise algorithm which is not helpful in all situations. The calculation of output texture from the input texture is an elegant approach as the algorithm adjusts to the size of the input and displays a good representation.
The use of Line Intergral Convolution algorithm in generating 3-d and 2-d images can be used in various important fields for further extraction of details from the images as this produces a better representation and images can be enhanced and better results obtained with this algorithm. The use of continuous directional motion filters in a local area can be used to implement simple and efficient processing systems.

Reaction: A survey of algorithms for volume visualization

The beginning of the paper starts with an affirmative quote of predicting the importance of volume rendering. The author clearly explains what and from the volume data is collected. This is a good precedence which actually gives us a good overview of the problem that is being addressed. As the paper is about the survey of algorithm , even the marching cubes is mentioned. Here we actually get to see other algorithms as compared to only marching cubes.

Reaction: Over two decades of Integration Based Geometric Flow Visualization

This paper provides an exhaustive information on different approaches used over the past two decades in using different visualization techniques for geometry based flow visualizations. This is an interesting read because it also gives an insight into how different challenges in the field were tackled while trying to find the best suitable visualization technique for flow simulations.
The broad level of categorization of vector field visualization approaches was a key thing in trying to provide information in a more meaningful way with the best possible terminology that they could have used. Explanation of the mathematical models and their structures giving the reasoning behind the visualization techniques used provides researchers to save time. The paper holds key in providing visualization techniques for very large data sets in real time applications. Most of the flow based visualization techniques quoted are already available as commercial products. With the ever increasing usage of GPUs for high end computation problems, the data sets are going to grow to larger within the same time. The paper stresses on the importance of trying to build better visualization methods for such systems as the current seeding algorithms provide cluttered visualizations.

Reaction: Over Two Decades of Integration-Based, Geometric Flow Visualization

The paper shows how flow visualization is a branch of scientific visualization.The paper presents the challenges in flow visualization along with statistic tables too.

The paper clearly presents the challenges seen in flow visualization for example no lag when a display is in from of us.Various categories of these have been explained too.

The paper was informative in knowing about flow visualization.

Reaction : Marching cubes: A high resolution 3D surface construction algorithm

Marching cubes is a new algorithm for 3D surface construction,
complements 2D CT, MR, and SPECT data by giving
physicians 3D views of the anatomy.

The steps listed which actually gives a complete overview of the exact steps reading four slices into the memory and finally outputting the triangle vertices and vertex normals, it the best part of the paper.

Another interesting part was the application of visualization into medical scanning and related fields.

Reaction: Over Two Decades of Integration Based Geometric Flow Visualization

In this paper the authors present us with a new type of visualization called Flow Viz which is related to the fluid mechanics and how to get valuable data out of those flows. It was good to read after the previous paper on LIC which also talked about the concept of fluid mechanics.
This paper was pretty good to read with authors explaining very clearly that numerous techniques for visualizations are available today and it is to the user's preference that which technique to choose to. The paper is well written and well compiled.
There are challenges that flow viz faces and one of them is large set of data. I agree with the authors regarding the seeding problem. A big thanks to the authors for presenting this idea and making us aware of this type of visualization.

Reaction: Imaging Vector Fields Using Line Integral Convolution

I was not familiar with vector field and this one was a good but a tuff read for me. Interesting to read about the LIC - Line Integral Convolution for imaging 2D and 3D vector fields. I had studied line drawing algorithm - DDA in undergrad and it was a good refresher here.
A good example of the photograph of flowers processed using the LIC and the blurring effect. Also in the implementation sections it was interesting to know how this LIC technique applies to the maps with the example of wind velocity visualization.

Reaction: Over Two Decades of Integration-Based, Geometric Flow Visualization

This paper starts off by stating some of the challenges that comes with flow visualization. It then moves on and describes the various flow visualization techniques and their respective pros and cons. One criticism that I do have is that the math embedded in this paper is making the paper harder to comprehend. Without a proper 3D or math background, it becomes increasingly difficult to follow the paper as it progresses. While the ample pictures in the paper is interesting and eye catching, it also does not help me understand the content.

Reaction: A Survey of Algorithms for Volume Visualization

This paper provides a broad overview of different volume rendering techniques and their tradeoffs. I find the organization of this paper to be particularly effective in helping me understand the content. The breakdown of data collection, data classification and rendering steps helps me comprehend the steps necessary to execute volume rendering. The comparison between different techniques is also very helpful in determing which is suited for my group project. Overall a very informative and useful paper.

Reaction: A Survey of Algorithms for Volume Visualization

In this paper the authors have presented a set of algorithms for the volume visualization. Good to read the five foundation algorithms. It has been 20 years since 1991 and we can see volume rendering and its increasing widespread use due to high performance processors and graphics. Did not really understand what the authors want to convey via photorealism. "jury is still out on this decision" doesn't conclude the discussion.
Future work is discussed at the end of the paper. It says that volume viz can be adapted by network based environment and parallel processing. Could we get the current results as what is the state of art technology used for volume rendering?

Reaction: A Survey of Algorithms for Volume Visualization

The paper begins with a list of applications of volume visualization and and the goals of a good volume visualization technique. The goals are simple and well stated that it should allow for quick data manipulation and re-rendering. Then they talk about the types of data available for volume visualizations. They point out that volume visualization data is mostly in vector form. Then there is some brief discussion on each of the underlying concepts used in volume visualization algorithms. This is very important as it helps understand the different approaches better.

The paper describes the different steps that are taken in usual volume visualization algorithms. They have also discussed the different algorithms. They have given pictures for different algorithms this is very helpful to understand the what will be the end result of these algorithms. They mention the trade offs taken in each algorithm and this explains that there is no one algorithm that fits all. In the summary they have revealed what is still left desirable in these algorithms.

Reaction: Marching cubes: A High Resolution 3D Surface Construction Algorithm

First thing that caught my eye is the date when this paper was written in - 1987 (which coincidentally is my birth year). Things have indeed changed since then but was a good paper to read. The authors have talked about the marching cube algorithm using the techniques that I am familiar with like the divide and conquer algorithm. I do not have any medical background so it was good to have an information flow in this domain. Good to know how technology is woven with medical domain.
Implementation at that period of time was done on C and Sun workstations as the authors mention. Wonder they will perform on todays multicore processors and high level languages. The diagrams not too clear e.g. cut MR brain but are a good attempt considering that the paper was published in 1987.

Reaction: A Survey of Algorithms for Volume Visualization

This paper discusses an abstraction of how the volume visualization techniques are applied without delving into the implementation specifics of these techniques.

The paper gives some interesting volume characteristics in the form of voxels, grids and lattices. After learning these characteristics, we understand volume visualization is not an easy thing and requires interpolation and sampling and subsequently checking their reliability. The paper also covers some of the visualization methods and data classification techniques to make the visualization more meaningful and highlight the important aspects. The article makes us aware that volume visualization consists of many levels of data processing.

The different visualization algorithms alongwith their advantages and disadvantages, requirements are specified. I found the most interesting algorithm to bet he ray-casting method because it works at the pixel level and can be made parallel.
The paper finally emphasizes the importance of correct volume rendering because it has important applications in the medical field.

Overall, this paper lends generality by covering procedures, heuristics, algorithms, data and processing methods.

Reaction: Marching cubes, A high resolution 3D surface construction algorithm

The paper talks about constructing a 3D interpolation from a set of 2D slices as input. The authors of the paper explain the problem in a simple manner. The authors discuss the related algorithms too and groups them based on the main idea behind these algorithms. Inter slice connectivity, surface location and surface gradient are the main features to be extracted from the slices. They point out that other algorithms fail to do one or more of these tasks in a more accurate manner. Keeping this as the basis they have proposed the marching cubes algorithm for the same. They have explained the idea of marching cubes very well. They have also reduced the number of possibilities of intersection of a surface with the cube from 256 to 14. They could have been a little more elaborate on how this was achieved. I think reducing the computation so much and also using advanced techniques to analyze and represent the gradient makes the approach really good. The results seem impressive however a side by side comparison with the results of other techniques would have helped in understanding the effect very well.

Reaction:A Survey of Algorithms for Volume Visualization

As the paper suggests it reads about volume visualization. It presents different types of volume visualizations algorithms and spots differences between them also.

Volume visualizations are used to create images from scalar and vector representations.
The paper explains many of the fundamentals of volume visualizations. It says how many of the problems are still unsolved by scientists and how certain expert instruments are needed to have better results,

Reaction:Marching Cubes: A High Resolution 3D Surface Construction Algorithm

The author presents a paper which gives explanation of the algorithm called marching cubes.It was interesting to know how polygonal representations could be done from 3D array data.

Te author did try comparing this algorithm to other ones too but that really dint help me understand the differences very clearly.

The research images of the 3d that the author presented helped in understanding the benefits of 3D over 2D.

Overall it was an easy paper to read due to he examples and the link of examples to medical field made it an interesting read for me.

Reaction: Over Two Decades of Integration-Based, Geometric Flow Visualization

I feel the authors present a valid discussion of the number of visualization techniques that are available today and how non-trivial it has become to select a technique based on user data. There are many factors which are to be taken into consideration especially what the users preferences are. This paper is well presented with each sub heading explaining the details needed to understand the following section. The authors have done a good job by making sure even those who are new to flow viz. can understand what the details of it.

The challenges in flow viz. are easy to understand and the problem with large data sets in not only limited to flow viz. but I believe it can be applied to any visualization technique. The user will certainly not want a delay to occur before his viz. is presented to him. The seeding problem however can be limited only to the geometric viz. as presented by the authors. I however do not completely agree with the authors view that perception is only a problem in 3D and 4D fields because it can be a problem in 2D fields too when we have multiple overlaps. CPU and processing power problem is again generic to all viz. and not only to geometric or flow viz.

This paper has also taught me the techniques of flow viz. which I did not know before. The tabular column presented aids to the ease of reading and understanding of the material. Overall the paper is an enjoyable read and there is a good flow of information from one topic to the other. I can most certainly say that it is one of the well documented research papers that I have read.

Reaction: Over Two Decades of Integration-Based Geometric Flow Visualization

Flow visualization is, let’s say a branch of visualization which is related to fluid mechanics and how to make flow patterns visible in them to get valuable information from them. The section of this paper that was intent of reading was very informative and definitive coming up with lots of important definitions. What I inferred from the reading was there are challenges faced in flow viz like
  • Tremendous volume of datasets – Since complex simulations are involved it produces huge amount and data and handling it efficiently becomes an issue.
  • Interaction, seeding and placement – Seeding used to place objects within data domain is real challenge
  • CPU processing time to handle this volume of data
  • Perception – this is like how to visualize 2D and 3D velocity fields and at the same time visualize multi-variate data sets.
Also I learnt about four categories of flow visualization mainly – direct, dense textured based, integration-based geometry and feature-based flow visualization
I appreciate the effort taken by authors in classifying the work done in this field over the past decades in the table they created. Overall structuring of paper was very informative for someone who had no clue before the paper about flow viz.

Reaction: Imaging vector fields using line integral convolution

This is an impressive paper dealing with Image processing techniques. I have learned about the techniques which are used to generate textures. One interesting point that I have noted is how DDA is inaccurate in making the assumption that a local vector field can be approximated by a straight line. A good explanation of filters is also provided, especially how LIC can be used with a combination of another filter to remove aliasing. It certainly makes sense to use LIC for accuracy because of its efficiency and the extra operation of removing the aliasing comes at no additional cost.

I feel that LIC performs much better when used along with other algorithms than acting stand alone, this is because LIC can take advantage of the post-processing results of other algorithms and act upon them to produce highly filtered results. There is also a striking similarity that I noticed between the results produced by LIC and the photo editing software's we see on mobile phones today. This I guess shows the importance of LIC and how varied its applications can be.

I like the way the authors have paved the path for future work by describing about the varied applications of LIC, specifically the one which talks about how LIC can be used to deblur images which have been blurred by a moving CCD cameras. This paper is a good read for researchers and students who are interested in the field of image processing. I would certainly look into some of the references provided in the paper.

Reaction: Imaging vector fields using line integral convolution

This paper reminded me of my undergrad course called Image Processing which had sinusoidal equations, normalization, convolution, pre and post processing etc. LIC provides a generalized imaging method for 2 and 3-D vector fields. What was a striking feature of the paper was that it tried to provide a new group of continuous motion filters to indicate directional flow. The example showing the blurred motion of waving hand left a hard impression on me and I think these techniques must be used in fast moving vectors like taking photo shoot of racing cars so on and so forth. By the look of it, the algorithm looks simple to implement and less taxing on the system. Using DDA vs LIC has its own performance consideration and tradeoff as former if faster but latter gives better quality, which one to use I think depends on a given user and given scenario.

Reaction: Imaging Vector Fields using Line Integral Convolution

This paper discusses about the importance of Imaging Vector fields. Imaging vector fields has applications in science, art, image processing and special effects. It is interesting to know that Imaging vector fields have applications in artist domains.

The author intelligently gave us the required background by explaining DDA convolution method and its drawbacks, which helped us understanding the importance of LIC and its improvements better. DDA is much more efficient in terms of Cells processed per second. But, LIC is much accurate and there is an improvement in quality. So, there will be a tradeoff in using these methods.

I think a parallel implementation of LIC will prove to be both fast and accurate, and will be an excellent way of Imaging vector fields. The authors have mentioned that LIC algorithm is inherently parallel. So, there should not be any problem in implementing the parallelized version of LIC.

Overall, the paper is an excellent read. The concepts are explained quite clearly with the help of before and after images.

Reaction: Over Two Decades of Integration-Based, Geometric Flow Visualization

This paper discusses the advancements in the area of geometric flow visualization over the last two decades. Flow visualization is one of the classic branches of scientific visualization. The authors of the paper exercised lot of care in carefully defining technical jargon related to flow visulization. The presentation of the material is quite clear and the explanation is totally based on images they have presented.

There is no wonder that, a lot of progress has been made in this area. But, the challenges of flow visualization have just took a different direction now. It is interesting to know that, specialized vector field compression methods in flow visualization is actively being pursued on the GPU.

Currently, there is a lot of research happening in optimizing the methods of flow visualization on nVidia CUDA GPU architectures. The new nVidia Tegra processor chip has all these encoded in its hardware to give the users ultimate gaming and visualizing experience.

Overall, the paper is very informative with lot of images clearly explaining the methods.

Reaction: Over Two Decade of Integration based Geometric Flow Visualization

This paper is an unaccounted survey for the different approaches used in geometry based flow visualizations. The read gives a thorough understanding of different challenges in visualization. The whole idea of trying to tackle the challenge of presenting a better flow visualizations for such large data samples in the associated fields with categorization of vector field approaches was a significant effort by the authors. The efforts put in building mathematical models for visualization techniques was key. The presented models have been used in the most current products that are of great commercial value. With modern usage of GPUs for many purposes apart from them being used for graphical applications in their early days, the data samples that need to be visualized have and will grow in an unexpected fashion and the authors stress on the fact that better algorithms have to be designed to be abreast with the architectures available, for the current seedings show that the requirement.

Reaction: Imaging vector fields using line integral convolution

This paper is significant as image vector fields has applications in science, arts, image processing and various special effects. Algorithms that can image directional information should have wide applications across scientific and artistic domains and would possess a number of desirable properties such as accuracy, locality of calculation, simplicity, controllability and generality. Line integral convolution is one of a kind. LIC is a technique that can image dense vector fields and is independent of both predefined sampling placement constraints and texture generation techniques. At the same time it can work in two/three dimensions. The comparison of the LIC algorithm with the by then present DDA algorithm shows its significance and effectiveness. The mathematical depiction of the model helps analyze the algorithm in a very lucid fashion. The LIC algorithm's success depends on the shape of the filter used for defining contrast of the system/ texture of the system. The output of the algorithm is very useful and can be operated in any way by defining various normalization values for brightness and contrast values of resultant image.

Reaction: A survey of algorithms for volume visualization

This paper is an old (two decades) survey of the volume visualization inception in the field of scientific visualization. Volume visualization is used for creation of high quality images from scalar and vector datasets defined on multidimensional grids. The paper talks of the algorithms that were used during nascent stages of volume visualization of data samples. The paper stresses on the importance of animation in volume visualization. The survey does not talk about the methods of handling Non_cartesian data. The paper starts off with explanation of different characteristics that the data samples can have by talking about their data and volume characteristics and explain the common steps taken in volume visualization algorithms. The volume visualization algorithms are divided into surface fitting and direct volume rendering and the direct volume rendering models are further divided into projection and image-order methods. The paper clearly explains the algorithms of all the specified in a succinct fashion and quotes them appropriately.

Reaction: Marching Cubes - A High Resolution 3d Surface Construction Algorithm

This paper allows us to come up with a mathematical model in deriving constant density surface from a 3d array of data running the simulations for every surface. This is achieved by creating a triangle mesh that approximates the iso-surface and by calculating the normals to the surface at each vertex of the triangle. The algorithm tries to locate the surface in a cube of eight pixels and calculate their normals. This process is repeated for all the cubes that can be formed. Surface intersection in a cube is done by assigning codes to the vertices outside/inside the surfaces. Triangulation is used to find the ways a surface may intersect the cube to patterns which are more convenient to be resolved. This paper provides with a very simple rendering technique and also an efficient way of manipulating data. This would also work for large amounts of data in high resolution images but the system can get really complex.

Reaction: A Survey of Algorithms for Volume Visualization

This paper essentially describes five of the most commonly used fundamental volume visualization algorithms and their close relatives. Descriptions of the algorithms include: what type of data they best handle, their advantages and disadvantages, a rough idea of how well they parallelize, their space and time requirements, a few possible optimizations and enhancements.

This paper answers several questions I had when I was reading the paper about "Marching cubes". In the Future work section, much was mentioned about no knowledge about visualizing volume data that is vector, multi-modal, multi-variate,higher dimensional and non-Cartesian. Also hardware implementations of DVR and SF algorithms should be readily available now. Since, the paper is written in the year 1992, I am curious what are the developments in the area of Volume visualization in these 10 years, and I ask the class to let me know if there are any set of papers I could read to understand the advancements.

Reaction: Marching cubes, A high resolution 3D surface construction algorithm

This paper presents an algorithm, called "Marching cubes" that creates triangle models of constant density surfaces from 3D medical data. The algorithm uses a divide-and-conquer approach to generate interslice connectivity, and further create a case table that defines triangle topology.

The idea is to extract a polygonal mesh of an isosurface from a 3-D volume. The algorithm works only with cubic-cells(voxels). The algorithm falls under Isosurface extraction category which still remains one of the most commonly used visualization methods. I am wondering if there are any other interesting and highly used visualization methods existing today. As the paper was published in 1987, I am just curious had there been any developments in this area.

Moreover, the paper mentions that other methods introduce artifacts, whereas the "Marching cubes" algorithm produces models with unprecedented detail. Though the algorithm is explained clearly, nothing is really mentioned about the complexity of the algorithm, and it was not compared with other extraction algorithms in terms of complexity.

Reaction: A survey of algorithms for volume visualization

I felt this paper was initially repetition of previous paper but later realized that it has made an interesting attempt to compare various algorithms and what was appealing was the way in which problems in volume viz. were placed with some ways to tackle them. Lack of images was making this paper hard read but I was fascinated that lot has been done to visualize data but same effort is needed to device ways to tackle issues faced every now and then. So much to say, the paper is good read for one can choose a algorithm or as per to say a method that is helping visualizations on demand kind of a thing.

Reaction: A high resolution 3D surface algorithm

The paper describes about an algorithm called “marching cubes” which sort of creates a triangular models of uniform density from 3D medical data. WitI liked the use of divide and conquer approach taken by paper to develop the topology. The steps right from using the algorithm, interpolation gradience is all well defined in the paper, but I was not on the same page as the authors where they mentioned the use of using triangulated use of cubes which I felt was too small of a number.

Overall I like the presentation of the paper, though I must agree was looking from retro days,and there was lot to learn if you come from non CG background like me. The authors must be applauded for their work in time frame in which this paper was written when CG wasnt as advanced as it is now or even five years before.

Reaction: A Survey of Algorithms for Volume Visualization

Another very good read for knowing the pointers to understanding volume visualization, the paper does a decent job in presenting to the user the algorithms used in volume visualization. I agree with the authors that volume visualization must offer quick data manipulation because it is important the user is presented the result of his selection parameters with as few clicks as possible. It is not possible to favor one algorithm at the cost of another, different algorithms have applications in different fields. If we look at the algorithms that are presented, none of them can most certainly be applicable to large dimensional data even though they work with large data sets.

The paper gets somewhat complicated as the content proceeds because this paper is purely based on algorithms used in computer graphics. There will definitely be many unanswered questions by the end of the paper because of the terminology used. I personally do not understand the color codes used for CT data, are these standards? If yes, then over a period of many years with color improvements the standards should have most certainly changed. Also, I feel the authors could have presented a more deeper insight into one of the algorithms so that it can be seen as an attempt to solve one of the many problems which are yet to be solved as described in the summary section. Even though the authors claim that the presented material is just an introduction to the algorithms used in volume visualization, the terminology could have been explained better.

Tuesday, October 4, 2011

Reaction: Marching cubes: A high resolution 3D surface construction algorithm

I found the paper to be a mix of mathematical analysis combined with the power of computer graphics by taking the medical field as a base set. It was very interesting to know how CT, MR and SPECT work and what is the difference between the three of them. The authors clearly state why their algorithm is superior when compared to the ones that previously existed, this I feel is important because the reader might keep pondering about the difference between the old and algorithm presented in the paper and hence removes the confusion. I however strongly argue about the validity of the statement which the authors have made which talks about "useful information" when relating to old algorithms. I think the validity of any information can only be based on what the experiment is trying to accomplish and its set of related inputs.

Looking at the fact that the paper was published in 1987, the authors have done a fantastic job by coming up with the algorithm and a working implementation which runs the marching cubes algorithm on the data produced by the CT, MR and SPECT. I seriously wonder how advanced was the field of 3D anatomy at that time because the images shown in the results section on the paper are very impressive. Also, one more point which I would like to make on the algorithm is the amount of space complexity and the processing time it takes and whether the algorithm is still used as a basis on which research work is done today.

Overall, this is a must read paper especially for the curious who would like to know how 3D modelling was done back in those days and how efficient it was (images given in the paper show that it infact did a pretty good job!).

Reaction: A Survey of Algorithms for volume Visualization

The paper is an excellent understanding of how volumetric datasets can be visualized in realtime where present day datasets are very huge and optimistic analysis is required to display them with tight time constraints. I think data classification and viewing and shading are the two important factors that influence the visual perception of the data and these are to be chosen suitably to get a better understanding of the data. The contour connecting algorithm seems simple, but there is a problem when there exists multiple triangles or contours formed in each data slice and there will be a confusion on which data slice to connect.

I think the marching cubes algorithm presented in the paper is an optimum way of representing the volumetric data and further improvements like dividing cubes can be used to avoid ambiguities and the gradient calculations also preserve small features of the data opposed to the opaque cubes method. So different algorithms can be used in different types of applications like opaque cubes can be used where we need to represent large sets of data and neglecting small features doesn't matter or we can use splatting technique where high quality images are needed and attention to detail is important. So, the paper helps a lot in choosing a proper algorithm that helps in visualizing data based on need.

Reaction : Over Two Decades of Integration-Based, Geometric Flow Visualization

The first section in the paper discusses more about the flow visualization techniques. With the improved power of infrastructure like GPU they have achieved a very wonderful result in visualizing them. The author goes through the major challenges faced in the paper like large data sets,interaction ,computation time, perception , classification etc. These were some of the question I had raised while reading the previous papers and so I can see that the paper has given a good justice to go through all aspects of volume visualization and info viz in general.

I would have liked if the wind direction imagery had some influence of LIC , as it would have given more reality to the viz. The paper gave insight into some of the terminologies like streamline and critical point. The study for reducing the occlusion problems and possibility of representing these in 3D would be some future work which could be done on the CFD area. So that it would help simulating fluid motion in medical, mechanical or aviation fields.

Reaction: Imaging vector fields using line integral convolution

Like the 'geometric flow visualization' paper, it is easier when the reader has background on the vector field domain. Comparing figure 2 and 4 shows the difference and advantage of LIC over DDA convolution. The circular and turbulent fluid dynamics shows clearer circular shapes and edges. Figure 12 which shows a painted version is interesting. Figure 13 which shows motion blurring displays the direction of the wave of the arm which is informative. A lot of image examples helps understand the purpose of the technique. One thing that I find (because of the performance) is that not only this technique but the previous spatial visualization methods depend or suggest parallel implementation.
In normalization, it seems that they have discovered that constant kernel normalization highlights singularities. However, in figure 8 about white noise convolved with fluid dynamics vector field using different types of normalization, is the bottom image a result that people are more interested than the top image?
There is a performance and quality trade-off between DDA and LIC. The CPS(cells processed per second) is 10 times faster for DDA than LIC. However, LIC has better quality. Which one does the user prefer?

Reaction: Imaging Vector Fields Using Line Integral Convolution

Line integral convolution represents a new an general method for imaging two and three dimensional vector fields. Using vector fields to draw imagery was an interesting read. The picture of the flower , wind velocity and fluid dynamics  were good examples to understand the  concepts of DDA and LIC. The author himself says a fourth order Runge Kutta method could produce differing or improved results. So it would be nice if the paper could have gone more in detail to all the possibilities of the mathematical methods possible.

It was interesting to know that visualization was interest of study to fields like fluid dynamics. Hopefully we would be able to see more accurate and better algorithms and deblurring techniques when research goes further in future.

Reaction: Imaging Vector Fields Using Line Integral Convolution

        This paper talks about Line Integral Convolution, A method for imaging two and three dimensional vector fields.  The LIC Algorithm takes an input image , vector field and generates an output image by filtering the input image along local stream lines defined by the vector field . "The LIC implementation is a module in a data flow system like that found in a number of public domain and commercial products. This implementation allows for rapid exploration of various combinations of operators. The algorithm can be used as a data operator in conjunction with other operators. Specifically, both the texture and the vector field can be preprocessed and combined with post processing on the output image. "
    It talks about DDA covolutions a generalization of traditional line drawing techniques and the spatial convolution algorithms. This summarizes LIC and also its applications.

Reaction:A Survey of Algorithms for Volume Visualization

    This paper is what its title says. It gives an introduction to various algorithms used for volume visualization."Volume visualization is used to create images from scalar and vector datasets defined on multiple dimensional grids, i.e., it is the process of projecting a multidimensional (usually 3D) dataset onto a 2D image plane to gain an understanding of the structure contained within the data."2-D visualization is hard, but 3D visualization is much harder. CT scan, MR scan are used for this 3D visualization. It talks about terminologies like vortex,cells. It concludes by stating problems like multivariate,non-Cartesian grids which are yet to be solved. This paper is hard for me to understand as I have no background in CG.

Reaction: Marching cubes: A high resolution 3D surface construction algorithm

Nice diagram about the flow for medical algorithms and the comparison among CT, MRI, and SPECT. Cutting the model using boolean operations is a nice and easy way to manipulate the visual object.
The result says that execution time depend on the number of surfaces and resolution. It took 30 minutes for a 260x260x93 CT data. Just out of curiosity, how long would marching cube take with the current hardware environment?
Is there advantage (or disadvantage) looking up the precalculated table in this algorithm? For example, if the viewpoint changes, it needs to recalculate the table which may negatively affect when user interaction is involved. Even if the object rotates, the table needs to be modified, right?

Reaction: A survey of algorithms for volume visualization

Interesting what Kajiuya had mentioned in 1991 that all rendering will be volume rendering. The fast improvement of hardware has supported a lot for volume rendering as research with GPU has increased. The paragraph about photorealism is located at a sudden and odd place, I think. I haven't seen a lot of photorealism applied to volume rendering but wonder if it is popular these days and in what kind of applications, if any. Considering the hardware support these days, most of the methods are parallelizable which seems plausible to make use of the introduced algorithms with GPU.

The paper says, creating static images from volumes of vector dta is an unsolved problem back in 1992. Is it still an unsolved problem?
It seems volume rendering needs pre-processing stage from the initial step of data acquisition and it is applied to every slice.
Is pre-processing of data required for all volume rendering?
What happens if there are negative surface pieces using SF methods? Would there be a hole after rendering the dataset?
Color code used for CT data, for example, would be bone-white/opaque, muscle-red/semi transparent, fat-beige/mostly transparent.But is there a standard for color coding or is it based on user selection?
What is false positive? What is negative triangles?
The authors talk about ethical issues and standard means of validating algorithms. Are there validating algorithms now?

Reaction: A Survey of Algorithms for Volume Visualization

The paper was a good read as it contained consolidated algorithms present in volume visualizations. It discusses about surface fitting and direct volume rendering methods. and the importance and difficulties of data classification. Data classification in 2D itself is difficult so we can imagine how difficult it would be in 3D. When they say color and opacity are one of the major factors in classification I guess the way people perceive things also be taken into consideration.

As the paper has mentioned this paper was not able to conclude problems of handling non cartesian girds, multivariate and higher dimensional data. A better classification algorithm also should be figured out. As human minds find it difficult to categorize the volumes just by seeing it , it becomes challenging to visualize the data using some animations and easy to use interfaces. So further study and detailed expiation of these basic points like data acquisition, slicing  , data set reconstruction and data classification would improvise volume visualizations in future.

Reaction:Imaging Vector Fields Using Line Integral Convolution

The application of imaging vector fields is not limited to just scientific applications. In this paper, an algorithm for imaging vector fields is presented.The algorithm filters an input image along local stream lines defined by an input vector field and generates an output image. The figure illustrating this algorithm helps in giving a clear picture to the reader. The DDA algorithm is also presented along with its shortcomings.Two vector fields are rendered using both LIC and DDA algorithms. It is interesting to see how certain details are not present or incorrectly rendered in the image generated using DDA.

The paper shows how the LIC algorithm can be used in post processing to generate motion blur. As the author suggests it would be worthwhile to see how future research work focusing on the reverse of this operation i.e deblurring a blurred image shapes up. The paper had a lot of images which made it easier to understand the concepts and techniques involved. Also, the mathematical derivations and the variables involved were well explained.

Reaction : Over Two Decades of Integration-Based, Geometric Flow Visualization

This paper acts a pointer to different research areas depending on the reader's interest. The visuals provided in this paper are very interesting and serves as a backbone in understating the areas. The paper is based on geometric flow visualization. It is in fact flooded with the information regarding that.

The jargon's like streamline, streak line are difficult for a novice reader like me. Some sort of description or examples should have been give to those which would have helped me in understanding. The description of various techniques to perform geometric visualization is very impressive. I see that flow visualization has lots of practical usages.

The fact that the last research was done six years back in this field shocked me. It is interesting to know how the challenges like huge amount data can effect the visualization as well how algorithms try to overcome that issue. The differentiation between interactive and automatic seeding is well done. I like the way overall seeding algorithm is presented.

There has been a healthy discussion at every point of the paper. Overall the paper give a good insight about the research done in the last two decades in this field.

Reaction: Marching cubes: A high resolution 3D surface construction algorithm

The paper discusses about information flow of  3D medical algorithms and a details of Marching cubes algorithm. Since the flow of algorithm was discussed before the the marching cubes algorithm was explained it was more easy to understand what the author is going to achieve. Even though the paper was written 25 years back, it has covered a lot of important concepts which we still follow. The divide and conquer approach used here has improved the efficiency of the algorithm.

 For reducing errors they have reduced the 256 patterns to 14 basic patterns , permutation and combination of these patterns produce the 256 cases. Even now in graphics we follow the triangulation approach for developing 3D images, which marching cubes has used, so I found this paper very valuable in this new era. The authors have considered lot of factors in mind thats why I think they have chosen stable operating systems like VMS , VAX and Unix for the implementation of the algorithm.

The possibility of selecting wrong set of  three points for generating the image in this algorithm is a disadvantage  . We can see that there were future work done in this regard which gave rise to algorithms like marching tetrahedra algorithm which generates more triangles and which needs more memory.

Reaction: Imaging Vector Fields Using Line Integral Convolution

The paper mainly concentrates on the importance of imaging vector fields. The fact that special effects can be produce by using these techniques, it is worth the reading. The examples and the illustrations are so interesting as they try to create special effects on a normal image. Finally their approach boils down to the usage of many techniques that were defined before like texture generation. This paper also discusses about the influence of color, intensity, and direction in scientific visualization. This a very new concept to me, especially the direction.

The description of LIC algorithm, helped me correlating with its practical usage in photo editing software like Photoshop, premiere pro etc. I feel that some more stress should have been give on the influence that this algorithm makes in the medical field, which is a major area where visualization plays a huge role. The growth of this algorithm from DDA algorithm is well mentioned. The drawbacks of the DDA algorithm like loss of accuracy and how the new LIC algorithm solves are described in a very eloquent fashion.

Reaction: Over Two Decades of Integration-Based, Geometric Flow Visualization

This paper mainly talk's about Flow Visualization techniques development over last 20 years.It starting author has inform about challenges in flow visualization.I find paper informative and useful for newcomer in field as it provides covers complete details about flow visualization.Paper also contains classifications and terminologies of flow visualization which helps in understanding forthcoming techniques to visualize flow.

All the techniques whether 2D/3D or Surface based/Volumes based were defined very clearly.Paper contained lot of images which were helped to understand and visualize working and output of these techniques(as it was hard to understand just by write-up. Eg. Streamline seeding surface Fig 4).Author has suggested that there are less research on higher dimension and lot of researches are done and are ongoing on seeding and streamline. Paper is very absorbing(interesting) however length of paper impacts reading quality toward later end of paper.

Reactions: Imaging Vector Fields Using Line Integral Convolution

The paper attempt to describe a new technique to visualize flows. Unlike other techniques, this is capable of producing an entire image at every rendering step. Beyond this high level concept, very little of this paper was comprehensible. The technical background required is far too steep for me to able to grasp any detail involving this technique. Unfortunately, the images provided by the paper offer very little guidance.

Reaction: Marching Cubes

This paper does a particularly good job at describing the algorithm behind the rendering using the marching cubes technique. The 15 unique configuration of the cube presents a good walkthru for how a specific polygon can be rendered. The usage of MRI images is particularly useful because it correlates exactly to our course project.

Unfortunately, for someone who is new to 3D rendering, the dictions used by this paper makes it hard to follow at times.

Reaction: Marching Cubes: A High Resolution 3D Surface Construction Algorithm

The author aims at describing an algorithm where small cubes constitute 3D space and then interconnects the surface. It is a very advanced paper with respect to the time when it was written. The conclusion that this algorithm successfully helped generating 3D proves the statement have I made previously.The comparisons that the author makes between different algorithms should have been detailed further. For a novice like me in the field it becomes really difficult to understand how this algorithm is better than its counterparts.

The fourteen patterns were well described and the explanation is self contained. As we can see the simplicity of this algorithm, if we implement the same on the fast end machines, it will be of great use. The illustrations clearly describe the practical implementations of the algorithm. The flow of the algorithm is simple and straightforward. The usage of divide and conquer makes this algorithm outperform its counterparts. The paper should be read with the fact that animation is need to represent there visualizations clearly.

Reaction: A survey of Algorithms for Volume Visualization

This paper aims at presenting detailed view of important algorithms in the field which are like the backbone to it.I felt that at some points the description was missing details. For beginners in the field like me, it is difficult to understand some high level details. More illustrations would have solved this issue. But the paper was written in an orderly fashion, first by mentioning about data visualization, then by volume visualization and finally a detailed description of the algorithms.

The algorithms are explained at depth making intriguing arguments. The concept of using light reflecting materials intrigued me, but they may be carcinogenic agents. The other factors like how easy to implement, which algorithm to use at what situation, drawbacks, positives should have mentioned, which would have made the paper look more practical.It is nice that the author mentions about the applications of the volume visualization, which helps reader in understanding the usefulness of the theory.

Reaction: Imaging vector fields using line integral convolution.

In this paper author has stated that imaging vector fields are important to scientific visualization and artistic domain.Paper mainly describes Line Integral Convolution technique to visualize directional information using vector fields.Even though it was hard to get DDA convolution in first reading ,I find description of DDA convolution understandable with the help diagram there.I find description of LIC technique hard for naive like me to understand, moreover mathematical details provided in LIC technique requires prior knowledge and makes it cumbersome.
I find second last page of paper more informative, where author has described that DDA and LIC have distinct trade-off of performance and quality between them.LIC is a magnitude slower than DDA but DDA is inherently inaccurate.Paper also states very important fact that LIC algorithm can be easily generalized to 3 dimension.

Reaction: Imaging Vector Fields Using Line Integral Convolution

This paper on Line Integral Convolution for Imaging Vector Fields is mathematically oriented.  Local Field Behaviour and Line Convolution have been explicitly brought out through a set of mathematical equations involving pixels and vector field at lattice point.  The concepts are well illustrated with multilayer diagrams which enable easier understanding of the process of visualization by changing the values of the related parameter. Simulating a brush stroke makes interesting illustration creating an inquisitiveness to probe into the algorithm in the areas of vectors producing contours of soft hills and valleys.  How a filtered input image and vector field combinations create an output image along stream lines defined by vector field have been enunciated in the paper. This provides added directional information during image processing. 

Extensive mathematical derivations illustrated in this paper have to be supplemented with mathematical learning especially in the areas of ripple filter functions and their periodicity.