| Summary: | Texture image retrievals are gaining its popularity among the Computer Graphics committees in the World Wide Web (WWW). Many a times, graphic artists failed to retrieve their desire results as most texture image retrievals available are based on descriptive text. Exploiting on their artistic nature, user input is proposed in the paper to be based on sketch. Leveraging on Content Based Image Retrieval, we combined shape context techniques together with feature tracking techniques to search for similar patterns. Shape context is used as a primary contention to search for potential similar patterns while feature tracking technique is used for similarity measures to compute the ranking between texture images. The combined effort of both techniques yields a satisfactory result of 95% on a texture database filled with alphabetical patterns, with each texture image uniquely representing a letter.
|
|---|