A SVD Based Pattern Matching Approach for Color Image Retrieval

Abstract

To make the pattern matching more accurate, local pattern matching algorithms could be applied upon suitably decomposed images. In this paper, SVD (Singular Value Decomposition) technique followed by LBP (Local Binary Pattern) algorithm has been applied to the images in RGB color space. Feature vectors are obtained for individual component of RGB and then merged to get the final feature vector. CBIR performed upon these feature vectors yielded better results and values of precision, recall and f-score was found to be 56.08, 84.13 and 67.3 respectively for the technique LBP-S-RGB with an optimal bin-size of 256.