Using Color and Local Binary Patterns for texture Retrieval

Texture plays a crucial role to detect and recognize materials in realworld pictures. The choice of the visual descriptors that are the most appropriate for the detection and recognition tasks is generally a hard issue, due to the wide range of circumstances under which the imaged objects can appear. This work addresses the problem of the illuminant invariant texture retrieval, and it proposes two algorithms that combine illuminant invariant color information with the textural cues output by the local binary pattern (LBP) operator. The experiments, carried out on a public database, show that the joint use of color and texture features remarkably improves the retrieval performance of techniques based on LBP only.