Wednesday, October 5, 2011

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.