| Hi-index | 0.00 |
This paper presents a novel region merging segmentation method for color image based on color and texture distribution features. The segmentation strategy includes two phases. In the first phase, we select initial seed points for super pixels extraction in the texture energy image at average intervals. Then we implement pixels clustering to extract over segmentation regions at local areas using color and texture information. In the second phase, a hybird texture histograms is introduced to represent the local color distribution information of internal pixels in over segmentation regions. The region merging employs computing corresponding histograms, which are normalized into fixed bins. Experiment results on Berkeley Segmentation Dataset (BSD) demonstrated that the proposed segmented algorithm can achieve good applications on the nature images with complex textures.