论文部分内容阅读
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算法完成对灰度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割效果要远优于其他基于散射图的阈值化方法．
Histogram is the most commonly used method for representing the statistical distribution of gray levels in images. The two-dimensional histogram includes co-occurrence matrix, grayscale-boundary scatter, grayscale-average grayscale and so on. In this paper, we propose a method of image segmentation based on the unsupervised clustering of pixels using the Maximum A Posteriori Probability (MPA) algorithm, which uses SEM algorithm to complete the parameter estimation of the two-dimensional normal distribution of grayscale-average grayscale. The test results show that our algorithm performs well, especially for low-contrast low-quality images with shadows and heavy noises, which are far superior to other scatter-based thresholding methods.