A new unsupervised segmentation algorithm is proposed in this paper. In the algorithm

the texture is first divided into overlapping windows

and texture feature is derived from each of the windows. The gray level of the pixel corresponding to each window is replaced by its feature. Then the image is projected into a feature superplane. An edge detection algorithm is applied to the feature superplane to detect where the texture feature changes rapidly. In this way

the texturel boundaries in the image are detected. The effective texture feature is crucial for the segmentation accuracy.