Texture image retrieval and similarity matching

Texture is one of the most important visual characteristics, which play a very critical role in many tasks, ranging from remote sensing to medical imaging and CBIR. The texture analysis has a long history and how to extract texture feature efficiently and accurately is still an active subject of study in the field of image retrieval. June proposed a method that used two features, and one of them is the gray-level histogram based on each low frequency. The gray-level histogram does not take the spatial relationship of gray in an image into account. In this paper, we establish a new way that describes texture in terms of their orientations and original image gray distributions using geostat, which represents the global spatial relationship of color. The performance has raised about 4% than that of June´s method.