基于亮度保持S型函数的双直方图均衡方法
2021-02-28王罡白洁静张正军
王罡 白洁静 张正军






摘 要:传统直方图均衡在保持图像亮度,避免过增强方面表现不佳。为了克服上述缺点,提出了一种基于亮度保持S型函数变换的双直方图均衡方法。首先用输入图像的灰度均值对直方图进行分割,然后根据子直方图的信息自适应地调整S型函数的参数,将子直方图的灰度均值变为其不动点,同时调整S型函数曲线的倾斜程度,最后用调整参数后的S型函数代替子直方图的累积分布函数进行双直方图均衡。通过对USC-SIPI数据库中的灰度图像进行试验,该算法在亮度保持和图像保真方面均取得较好的效果。
关键词:图像增强;双直方图均衡;S型函数;亮度保持;峰值信噪比
中图分类号:TP391 文献标识码:A文章编号:2096-4706(2021)16-0073-06
Bi-Histogram Equalization Method Based on Brightness Preserving Sigmoid Function
WANG Gang1, BAI Jiejing1, ZHANG Zhengjun2
(1. Department of Quality Education, Nanjing Vocational College of Information Technology, Nanjing 210023, China; 2. College of Science, Nanjing University of Science and Technology, Nanjing 210094, China)
Abstract:Traditional histogram equalization performs poorly in maintaining image brightness and avoiding over enhancement. In order to overcome the above shortcomings, a bi-histogram equalization method based on the brightness preserving sigmoid function transformation is proposed. First, the gray mean of the input image is used to segment the histogram. Secondly, the parameters of the sigmoid function are adjusted adaptively according to the information of the sub-histogram, then the gray mean of the sub-histogram is changed to its fixed point, and the steepness of the sigmoid function curve is adjusted. Finally, the sigmoid function after adjusting the parameters is used to replace the cumulative distribution function of sub-histogram for bi-histogram equalization. By testing the grayscale images taken from the USC-SIPI database, the proposed method has achieved good effects in terms of brightness preserving and image fidelity.
Keywords: image enhancement; bi-histogram equalization; sigmoid function; brightness preserving; Peak Signal-to-Noise Ratio(PSNR)
0 引 言
图像增强是图像处理中的一个重要领域,其目的是提高图像质量,以改善图像的视觉效果或者增强图像的某些特定信息,来满足人眼视觉或者机器视觉的要求[1]。在众多图像增强算法中,直方图均衡由于其算法简单快速,并且可以相对有效地增强几乎所有类型图像[2],目前已成为最受欢迎的图像增强技术之一。直方图均衡运用输入图像的累积分布函数作为变换函数,对原始灰度级进行调整,扩展图像的动态范围,提高图像的整体对比度。然而该方法主要缺点是当输入图像直方图存在尖峰,直方图均衡会导致图像过度增强[3],而且运用累积分布函数作为变换函数可能会显著改变输入图像的亮度,不能满足电子产品的增强需求[4]。……
