基于灰度直方图单一图像噪声类型识别研究.docVIP

基于灰度直方图单一图像噪声类型识别研究 　　摘 要：图像噪声类型识别是抑噪方案研究的前提及噪声参数估计的基础，对后续的图像处理有着重要作用。利用含噪图像灰度直方图特性，对几种常见单一噪声类型进行识别。首先对这些噪声基于统计学的方法进行数学建模，获取不同类型噪声的随机矩阵，把这些噪声矩阵加载到灰度图像中；然后对图像中灰度等级相对一致的不连续区域分别采样并画出直方图，通过直方图中不同分量灰度值的分布及图形形状识别噪声类型。在前人的基础上增加了瑞利噪声、伽马噪声、指数噪声等类型识别，对图像灰度直方图绘制方法进行了改进，扩大了采样区域的可选范围。对同一加噪图片均匀不连续区域分别采样，提高了噪声类型识别准确度。 　　关键词：灰度直方图；噪声类型；灰度级；像素点；含噪图像 　　DOI：10.11907/rjdk.172618 　　中图分类号：TP317.4 　　文献标识码：A 文章编号：1672-7800（2018）004-0197-04 　　Abstract：The identification of the image noise type is not only a prerequisite for the study of the noise suppression scheme， but also plays an important role in the subsequent image processing， and it is the basis of the noise parameter estimation. The recognition of the noise type is more and more important. This paper identifies several common single noise types mainly based on the characteristics of gray histogram containing noise image. Firstly， the noise is based on the mathematical method of mathematical modeling， and the matrices of different types of noise are obtained. These noise matrices are loaded into the gray scale image， then the discontinuous regions of the gray scale of the image are sampled and their histograms are drawn separately. The different noise types are separated through the histogram of different components of the gray value of the distribution and graphical shape. In this paper， the method of Rayleigh noise， gamma noise， exponential noise and so on are improved on the basis of predecessors， and the method of drawing image histogram is improved. At the same time， the range of sampling area is expanded. Sampling of different discontinuous regions also improves the accuracy of noise type identification. 　　Key Words：gray histogram； noise type； gray level； pixel； noisy image 　　0 引言 　　图像是信息传输的常见形式之一，广泛应用于生产、医疗、航天和军事领域，以及微信、数字电视等生活当中。然而，图像在形成、传送、接收等过程中难免受到噪声污染。噪声湮没了原始图像结构和纹理，影响视觉质量及后续的图像处理[1]。因此，图像的去噪处理尤为重要。 　　目前，大多数噪声研究假设噪声为已知类型及已知参数，而在实际应用中图像噪声类型一般未知[2]。越来越多的研究成果显示，正确估计和利用图像噪声对图像去噪及后续处理有着重要的指导意义与使用价值。再好的噪声去除方法皆不可能适用于