New approach for color-texture segmentation of natural images based on the distribution evenness of colors
基于色彩均匀度的自然图像色彩—纹理分割方法

Many present color-texture segmentation methods are either too computing complicated or lack of smoothness in the final results. To solve these problems, a new approach for color-texture segmentation of natural images was proposed, which was based on the distribution evenness of colors and called ISBEC for short. Firstly, input image color was quantized. Then the resulted index map was used to perform simultaneous color segmentation and multi-scale texture analysis. Finally, these results were combined together, and over-segmented regions were merged. ISBEC can also be used to segment gray-scale images if they are converted to color ones in which three color components of each pixel are identical. The experimental results demonstrate the effectiveness of ISBEC for color-texture segmentation of natural images.