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?
0 comments:
Post a Comment