一种二维直方图阈值化图像分割的后处理方法  

A post-processing method of two-dimensional thresheolding image segmentation

作　　者：张爱华[1] 余胜生[1] 周敬利[1] 

机构地区：[1]华中科技大学计算科学与技术学院

出　　处：《华中科技大学学报（自然科学版）》2002年第10期59-61,共3页Journal of Huazhong University of Science and Technology(Natural Science Edition)

摘　　要：在灰度图像的二维阈值化分割方法中 ,存在部分可能的噪声或边缘点 ,而没有对其进行处理或处理很草率 .针对这个不足 ,提出了一种进一步将这些点归类、提高分割质量的后处理方法 .该方法通过考察这些点的 8 邻域或 1 6 邻域内的其他点的分布情况 ,比较这些点到相应的邻域中分属于目标和背景的两类点中心之间的距离 ,将这些点分割到距离较小的一类中 ,以达到将图像中的所有点进行归类的目的 .实验分析表明 ,这种后处理方法提高了分割的准确性 ,减少了分割错误 .According to the two-dimensional thresholding algorithm, there are some pixels (edge points or noises) which are unsegmented or processed curtly. A post-processing method of two-dimensional thresholding image segmentation is presented to process these pixels and improve the quality of segmentation. Before classing a point of these edges/noises, the distribution of pixels in its 8-neighbor or 16-neighbor is to be investigated and the distances between this point to the center of pixels of object and the point to the center of pixels of background in its neghbor are to be compared, as well as the points should be put into the closer class. The experiments have proved that the method improves the segmentation and reduces the possibility of mistakes in segmentating.

关 键 词：直方图 后处理方法 图像分割 二维阈值化 灰度图像 灰度值分布 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]