Abstract
A new image feature descriptor for content-based image retrieval is proposed, named as Square Texton Histogram (STH). STH is derived based on the correlation between texture orientation and color information. Based on julesz's texton theory Square Texton' templates are proposed for Image texture analysis. Texture Orientation is computed by using proposed multi texture orientation detector that incorporates horizontal, vertical and diagonal edges information. Features are extracted by correlating texture color and edge orientation by using 4-directional co-occurrence matrix while; the final set of features is obtained by histogram. To find similarity between query and target image, a weighted square-chord distance measure is proposed. The Proposed distance metric integrates the advantages of both bin-by-bin and weighted distance metrics. The proposed STH method is tested on standard dataset's that are extensively used in CBIR domain, such as Coral5K and Coral10K. STH has good discrimination power of primary visual features.