Texture image segmentation method by using pyramid linking and self-organizing neural network

This paper presents a segmentation method by using pyramid linking with supervised neural networks and unsupervised neural networks. This approach is applied to segment an random texture image into several homogeneous areas. In this method, it first segments the image at coarse resolution and then links the result to finer resolution until individual pixels are classified. The texture properties of the image are extracted by using a two-dimensional autoregressive method and fractal dimension, and then feature vector pyramids are built of reduced-resolution versions of these arrays. By using band-pass neural networks, the pyramid node linking process is applied. The resulting trees defines a segmentation of the original image. The validity of this method will be verified by several numerical examples.