基于增强型EM模型重叠区域图像分割算法

Based on the Enhanced EM Model Overlap Region Image Segmentation Algorithm

针对图像大规模重叠区域的有效分割一直是一个难题,传统的Log算子、Sobel算子、Canny算子以及梯度算子等算法解决大规模像素重叠问题时,模型会陷入不收敛的境地,导致分割效果较差,为了解决这样问题,提出一种增强型EM模型解决重叠区域图像分割的问题,利用curvelet变换在curvelet域内提取图像的边缘特征,并定位特征curvelet系数.通过增强特征curvelet系数,增强边沿特征对比性,分割多尺度多结构元素形态学检测的边缘图像,消除重叠带来的干扰.仿真实验结果表明:分割的边缘更为完整准确,取得了令人满意的效果.

In image large--scale overlap region of the effective segmentation is always a difficult problem, the traditional Log operator, Sobel operator, Canny operator and gradient operator and algorithms to solve large scale pixel overlap problem, model will in no convergence situation, causing segmentation effect is bad, in order to solve this problem, the paper proposes a kind of enhanced EM model is used to solve the problem of overlapping area image segmentation, use eurvelet transform in curvelet domain of images edge features, and positioning characteristic curvelet coefficient. Through the enhancement features curvelet coefficient, enhance the edge feature of kriging, multi--scale segmentation more structural elements morphological detection edge image, and eliminate the interference of overlapping brings. The simulation results show that the edge of the segmentation is more complete and accurate, and satisfactory results have been obtained.