Project Description
Develop tools for SAR image processing, include speckle removal and edge detection
First, we can implement the refined Lee filter for speckle removal. This filter will estimate the local variance of the pixel based on different edge orientation. The merit of this filter is that it can retain the edge information unlike other de-noising filters that may blur the edges. Also this filter is relatively easy to implement and runs very fast
Another more complex and maybe more efficient speckle removal method is based on wavelet transform. This method can also be applied to robust edge detection with the presence of speckle noise. The wavelet transform can figure out the places with high discontinuity, which implies the presence of an edge. To check whether this is edge or noise, we can check the neighborhood of this point along different directions. If along some directions the neighborhood also has high wavelet coefficient, it means this is edge not random noise. When we find all possible edges, we can apply shrinkage to the remaining coefficients to remove speckles.
Availability
UTC-05:00 Eastern time 9:00AM ~ 6:00PM
Schedule
| Week | Tasks to accomplish this week |
|---|---|
| Week 1 31-May-2010 12:00 UTC | Begin refined Lee filter implementation |
| Week 2 7-Jun-2010 12:00 UTC | Finish refined Lee filter implementation |
| Week 3 14-Jun-2010 12:00 UTC | Begin local variance based edge detection |
| Week 4 21-Jun-2010 12:00 UTC | Finish local variance based edge detection |
| Week 5 28-Jun-2010 12:00 UTC | Testing the edge detector algorithm and add threshold input dialog |
| Week 6 5-Jul-2010 12:00 UTC | Implement 2D wavelet transform |
| Week 7 12-Jul-2010 12:00 UTC | Implement 2D wavelet reconstruction for image, midterm report |
| Week 8 19-Jul-2010 12:00 UTC | Implement shift-invariant wavelet transform and inverse transform |
| Week 9 26-Jul-2010 12:00 UTC | Combine wavelet transform and soft threshold for speckle removal |
| Week 10 2-Aug-2010 12:00 UTC | Try to implement texture segmentation for SAR images |
| Week 11 9-Aug-2010 12:00 UTC |
Finish texture segmentation for SAR images |
| Week 12 16-Aug-2010 12:00 UTC |
Final report and submission |
Add Comment