Effects of Arbitrary-Shaped Regions on Texture Retrieval

Texture is a widely used feature in content- and region-based retrieval systems. Most studies focus on texture extraction over an entire image or use color-texture combinations for region-based retrieval. Less attention has been drawn towards texture extraction and its evaluation over arbitrary-shaped regions. In this paper frame-based and region-based texture extraction schemes are compared to evaluate the effect regionalization on the texture description and retrieval performance. Experiments are carried out using MUVIS framework, which allows us to test and compare several well-known texture descriptors on texture databases. The results indicate that significant degradations over the retrieval performance occur on such texture descriptors whenever applied on arbitrary-shaped regions.