Texture Segmentation Using Sequential Kernel Density Approximation

In this paper, we present a scheme for texture segmentation using sequential kernel density approximation. For sequential kernel density approximation, every texture region´s intensity distribution can be described to a mixture Gaussian model, and the number of mixture model need not be set in advance. Exploiting intensity distributions directly leads to a region based measure for well-suited texture discrimination. Together with the mean of image intensity, a novel texture discrimination method is obtained. A demonstration of the performance of the scheme in this paper is given in the scope of texture segmentation.