MULTI RESOLUTION FEATURE EMBEDDED LEVEL SET APPROACH FOR TEXTURE SEGMENTATION

The work comprising of unsupervised segmentation has been rigorous topic of research. The existing work towards discerning various texture regions is by extracting different texture regions based on exceptional characteristic descriptors. The work presented in the paper signifies an integrated approach towards multi resolution texture segmentation exploring the possible methods of two region textures based on feature extraction using suitable filters and further embedding thus obtained features into the level set algorithm for further segmentation. The major effort in the projected work is the capability to acquire comparable texture regions that combines the region based on active contours and texture features. The results on images with two textures from Brodatz database, and also custom designed texture pattern which are available in Microsoft office database having subjective boundary have been qualitatively presented and the results are validated through a parameter called Intersection over Union (IOU).