Improved multiscale matched filter for retina vessel segmentation using PSO algorithm

The concept of matched filter is widely used in the area of retina vessel segmentation. Multiscale matched filters have superior performance over single scale filters. The proposed approach makes use of the improved noise suppression features of multiscale filters. A major performance issue here is the determination of the right parameter values of the filter. The approach employs particle swarm optimization for finding the optimal filter parameters of the multiscale Gaussian matched filter for achieving improved accuracy of retina vessel segmentation. The proposed approach is tested on DRIVE and STARE retina database and obtained better results when compared to other available retina vessel segmentation algorithms.