A Color Based Image Segmentation and its Application to Text Segmentation

The goal of this article is twofold. First, it deals with color image segmentation in hue-saturation space. A model for circular data is provided by the vM-Gauss distribution, which is a joint distribution of von-Mises and Gaussian distributions. The mixture of W-Gauss distributions is used to model hue-saturation data. After segmentation, a post processing based on both spectral and spatial similarity of clusters is applied to separate such identifiable objects in the image. The results and comparisons are shown on Berkeley segmentation dataset. The problem of text extraction from a color image is taken as an application of the proposed method. We use a laboratory made text image dataset to test the method.