Texture image segmentation method based on multi-layer CNN

This paper presents a new texture feature extraction method called Simple texel scale feature (STSF) based on the scale and orientation information of texels. And a new texture image segmentation method based on binary image processing is introduced. The scale information of texels is extracted by comparing the gray value of two pixels. The position relation of these two pixels shows the frequency and the orientation feature of texels. Texel scab features can, be extracted by using different position, relation (distance and orientation) After obtaining texture feature images, we consider the texture image segmentation problem not as a pattern classification problem but several texture edges integration problems, which are simple binary valve lines processing problems like Holes filling, Lines thinning and shorting. A new kind of Multi-layer Cellular Neural Network (CNN) called MLCNN is proposed, and some MLCNNs are designed for these problems.