新的噪声污染灰度图像边缘检测统计方法

Novel statistical method for edge detection of noisy gray scale image

针对传统算子进行边缘检测时易丢失边缘信息和在非边缘处增强噪声的缺陷,提出一种基于非参数变点统计分析的噪声图像边缘检测方法,该统计方法不但不需要图像数字特征的任何先验信息,而且对噪声污染的图像不作任何滤波处理。实验结果表明,提出的算法优于Sobel算子,并能抑制信噪较低的高斯噪声和密度较高的椒盐噪声对分割结果的影响,是一种有效的噪声污染灰度图像边缘检测方法。

The traditional edge detection algorithms had the defect losing details of the edge and the defect enhancing noise in the non-edge, so this paper proposed a novel method of edge detection of noisy gray scale image which was based on the non- parametric change point statistic analysis. It not only minimised the need for a priori information about images, but also didn＇ t filter any noise as well. It was not only suitable for detecting the edge of images added Gaussfan white noise, but also for detecting the edge of images added salt and pepper noise. Experimental results show that the algorithm is superior to Sobel algorithm, and can suppress Gaussian white noise of low SNRT and salt and pepper noise of high density. In all, it is a valid method of edge detection for gray images contaminated by noise.