基于TSPL的图像椒盐噪声去除新算法  

TSPL based new algorithm for removing salt and pepper noise from images

作　　者：朱效丽[1,2] 李防震[1,2] 

机构地区：[1]山东财经大学计算机科学与技术学院,山东济南250014 [2]山东省数字媒体技术重点实验室,山东济南250014

出　　处：《现代电子技术》2013年第14期96-99,共4页Modern Electronics Technique

基　　金：国家自然科学基金(30600121);山东省优秀中青年科学家科研奖励基金(2007BS09002);山东省自然科学基金(ZR2009GQ015)

摘　　要：在提出的三子集划分的灰度图像分层表示(TSPL)算法的基础上,主要研究对常见的椒盐噪声的去除。TSPL算法是一种新型的数字图像变换方法,其核心思想是用一系列具有解析形式的灰度函数逼近不规则的原图像灰度函数。在此分别处理由TSPL算法生成的各层基函数图像,将其划分为A,B,C三个区并将每个像素点赋值为0,1,2;结合图像椒盐噪声的特点,利用投票策略处理每个像素点的8邻域,从而达到椒盐噪声去除的目的;最终通过TSPL算法对基函数重构来恢复原图像。采用在人类视觉系统的前提下提出的基于结构相似性的方法MSSIM算法作为图像质量评价的标准,实验结果表明,在主客观方面,该方法在去除噪声和保留图像细节方面明显优于传统的中值滤波方法。The removal of salt-and-pepper noise from the images is researched in this paper base on the triple-subset parti- tion based image layer-presentation （TSPL） algorithm proposed before. TSPL algorithm is a new image transformation method. The key concept of the algorithm is that the original image grayscale function, which is comparatively irregular, is approximated by a series of high-regular grayscale functions. The base functions generated by TSPL algorithm are processed respectively, divid- ing each of them to A, B and C partitions, and assigning a value of 0, 1 or 2 for each pixel. In combination with characteristics of salt-and-pepper noise, the voting strategy is adopted to process the 8-neighbor of each pixel for removing salt-and-pepper noise. Finally the primary function is reconstructed by TSPL algorithm to resume the original image. MSSIM algorithm based on human vision system and structural similarity is adopted as the standard for image quality evaluation. Experiment results show that, in both objective or subjective aspects, the new approach is much better than the traditional median filtering method, which can remain detailed information of original images while removing the noise.

关 键 词：TSPL算法 投票策略 椒盐噪声 中值滤波 MSSIM 

分 类 号：TN964-34[电子电信—信号与信息处理]