摘要
为了增强图像分割算法的鲁棒性,避免出现错误或间断的边缘轮廓曲线,获得准确的区域分割线,提出区域边界最优映射分割(ORBM)算法。该算法采用Gibbs分布定义区域分割模型,将多个颜色空间的不同边缘映射求平均值,用得到的边界最优映射确定邻域(相邻像素)的相互作用势函数,利用α-β交换算法求解标签参数空间上目标函数的局部极值并采取简单区域合并策略,获得准确、可靠的区域分割结果。将ORBM算法与几种经典的图像分割算法进行对比,实验结果显示该算法能够生成连续封闭的边界线,实现了图像多区域的正确分割,并且执行速度快、鲁棒性强。
In order to enhance the robustness of image segmentation approaches and avoid producing false or disconnected contours, this paper put forward the optimal region boundary map （ORBM） segmentation algorithm to attain accurate segmenta- tion lines. The algorithm defined the region segmentation model expressed by Gibbs distribution. It averaged the several edge maps based on Canny and K-means to obtain the optimal boundary map for six different color spaces, and computed the interac- tion potentials of adjacent pixels from the optimal boundary map. It solved the optimization of the objective function over the la- bel space by using c^q/3 swap algorithm. Finally, achieved the reliable and accurate segmentation results through merging the simple region. It compared ORBM algorithm with several state-of-art image segmentation algorithms. The experiments show that the proposed algorithm is efficient to image region segmentation by producing closed continuous boundary, and has excellent performance and robustness.
出处
《计算机应用研究》
CSCD
北大核心
2016年第1期307-310,共4页
Application Research of Computers
基金
山西省基础研究计划项目(青年)(2012021012-04)
太原科技大学博士科研启动基金资助项目(20132026)
太原科技大学校青年基金资助项目(20103003)
关键词
空间融合
边缘映射
区域分割
spatial integration
edge map
region segmentation
