Research on texture image feature extraction method

Abstract

In this paper, we give several classical feature extraction methods, including grayscale co-generation matrix, Gabor and wavelet transform features, and local binary pattern series features. We introduce the basic principles of these feature extraction algorithms and some derivative methods respectively. Finally, we analyze the advantages and disadvantages of the existing feature extraction methods: grayscale covariance matrix can analyze the arrangement rules of image texture and extract local spatial features of the image, filtering methods and local feature extraction methods are widely used, but the extracted features do not provide a good description of the image structure; and the multi-feature fusion operation brings huge computational effort. Therefore, the future developable directions are proposed based on the existing problems and difficulties in processing texture images.