A novel texture classification method using multi-directions main frequency center

This paper presents a novel texture classification method using multi-directions main frequency center. A texture can be viewed as an approximately period signal. Its main frequency center can characterize the periodicity features very well. For a given texture image, the main frequency centers in 5 directions are firstly calculated, which combine the average of gray level of the texture to form a 6 dimensions feature vector. Finally, the minimum distance classifier is used to classify the textures. A data set containing 16 kinds texture from Brodatz album is employed to test our method and encouraging experimental results have been obtained.