低质量图像基于散射图SEM估计的MAP像素聚类方法

直方图是最常用的图像灰度统计分布的表示方法，二维直方图包括共生矩阵，灰度－边界值散射图，灰度平均灰度散射图等，本文提出了一种采用ＳＥＭ算法完成对灰－平均灰度的二维正态分布假设的参数估计，然后采用最大后验概率（ＭＰＡ）准则进行像素无监督聚类的图像分割算法，测试结果显示，我们的算法性能良好，尤其是对低对比度，有阴影和重噪声的低质量图像的分效果远优于其他基于散射图的阈值化方法。

MAP Pixel Clustering Method Based on SEM Estimation of Scatter-Plot for Low Quality Image

Histogram is the most commonly used method to illustrate the grayscale statistical distribution of an image.2D histograms include cooccurrence matrix,(graylevel,edgevalue) scatter plot(graylevel,local average graylevel) scatter plot etc.In this paper,we present an image segmentation algorithm which uses SEM algorithm to implement parameters estimation of the hypothesis of 2D normal distribution about (graylevel,local average graylevel)scatter plot,and uses MAP criterion to achieve unsupervised pixel clustering.Compared with other thresholding methods based on scatter plot by applying them to various test images,our proposed algorithm demonstrates better performance,especially for the low quality images with low contrast,shading and heavy noise.