Using spectrum to extract texture feature

Extracting texture features is a key step in texture analysis. Various methods for texture analysis have been proposed (see Salari, E., Pattern Recognition, vol.28, no.12, 1993; Chen, J.L., Proc. ICASSP, vol.3, 1992). An efficient description of texture features is necessary, and the computational load is an important aspect. Recently, the texture spectrum method (see Wang, L. et al., Pattern Recognition, vol.25, no.3, 1992) has been developed. Compared with other methods, the concept of texture spectrum is clear, and its computational load is comparatively low. We studied texture spectrum, used it to extract texture features and obtained a more efficient description of texture. The paper introduces the principle of this method and gives test results. The results show the reasonableness of the new approach.