Abstract: First discussed the features of segmentation algorithm that considered the spatial information. It has been found that the better segmentation result was achieved in coarse fabric than small texture one. If the small texture fabric has many details in the image, the segmentation algorithm that considered only color space information could be selected. Based on the above discussion, we should classify the fabric before segmentation. Experiments proved that the texture information mainly existed in the 2 and 3 detail sub image. Based on the decomposition coefficient energy, can distinguish the classification of the fabric from the texture view. And the coarse level can be detected too by the energy ratio of 2 and 3 detail images.

Key words: fabric image, segmentation, mean shift, wavelet transformation