The paper researches texture extraction using wavelet transform. After introducing the wavelet transform and the texture analysis methods, the image is decomposed by wavelet transform, and the sub-images are gained. Secondly, the paper takes entropy and mean as texture parameter, so the texture image is an entropy or mean image. Finally, the image is classified by the spectral and texture information. The size of the texture calculating window and the treatment to the sub-image are researched in this paper. On condition that the spectral classification adding with texture feature, the precision will improve 4% averagely. Wavelet transform can decomposed image at several levels, so it can provide many information to classify and extract, which is helpful to those applications. Because of the texture window, texture image has fuzzy edge, it will lead to error for the image that have fine object or the area with different objects interleaved.