基于相似系数的颜色纹理特征在图像分割中的应用

Application of Color Texture Feature Based on Similarity Coefficient in Image Segmentation

如何在保证图像分割算法的鲁棒性与准确性的同时,增加其算法的自适应性是现在的热门问题,提出了一种基于相似系数的约束与局部直方图彩色纹理相结合的图像分割方法。首先,将图片进行颜色量化;然后,针对已量化的数量足够多的颜色代表进行有效地聚类;通过统计其圆形邻域内每个代表颜色的像素数从而得到局部直方图,将圆形邻域构建多尺度的金字塔后,在相应的像素处采用一个适当尺度的局部直方图作为颜色纹理的描述符对图像进行分割;然后,对边界附近的颜色纹理进行了相似系数的约束校正,并针对图片利用颜色纹理描述符号进行聚类方法下的初始分割;最后,为了得到准确且良好的分割结果,在得到具体边界的条件下进行区域融合,目的是将过度分割的区域进行合并。通过在彩色纹理的图像上进行大量实验表明:该方法在PRI与GCE的数据对比中比其他算法更加优秀,从而证明了在保证分割算法的鲁棒性与准确性的同时进一步实现了算法的自适应性。

How to ensure the robustness and accuracy of the image segmentation algorithm and increase the adaptability of the algorithm is a hot problem.An image segmentation method based on the constraint of similarity coefficient and local histogram color texture was proposed.Firstly,the image was color quantized.Then,effective clustering was carried out for enough color representatives that have been quantified,the local histogram was obtained by counting the number of pixels representing color in its circular neighborhood.After constructing a multi-scale pyramid in the circular neighborhood,an appropriate scale local histogram was used as the descriptor of color texture at the corresponding pixels to segment the image.Then,the constraint correction of similarity coefficient was carried out for the color texture near the boundary,and the color texture description symbol was used for the initial segmentation under the clustering method.Finally,in order to get accurate and good segmentation results,the region fusion was carried out under the condition of obtaining the specific boundary,in order to merge the over segmented regions.A large number of experiments on color texture images show that the method is better than other algorithms in the data comparison between PRI and GCE,which proves that the adaptability of the algorithm is further realized while ensuring the robustness and accuracy of the segmentation algorithm.