Comparative Study Of Gabor, Circular Gabor and DCT based Method With Contextual Information For Unsupervised Texture Segmentation

摘要

In this paper the concept of pixel wise feature extraction for texture segmentation has been extended to block wise feature extraction with the addition of contextual information. A comparative study has been made among different texture feature extraction techniques with this concept. Computational burden is small due to processing in non overlapping blocks. It has been observed that Circular Gabor filter followed by DCT gives the best performance. The results have been validated in presence of noise.