光学 精密工程, 2014, 22 (8): 2214, 网络出版: 2014-09-15
基于非下采样Contourlet变换系数直方图匹配的自适应图像增强
Adaptive image enhancement based on NSCT coefficient histogram matching
图像增强 图像去噪 非下采样Contourlet变换 系数直方图匹配 自适应图像增强 image enhancement image denoising NonSubsampled Contourlet Transform(NSCT) coefficient histogram matching adaptive image enhancement
摘要
由于非下采样Contourlet变换(NSCT)域图像增强方法需要手动调节参数,无法实现自适应增强, 本文将直方图均衡化和NSCT域增强相结合,提出了一种基于NSCT系数直方图匹配的自适应图像增强算法。该算法首先对低对比度含噪原图像进行直方图均衡化,然后对原图和直方图均衡化后的图像分别进行NSCT分解,得到低频子带系数和各高频方向子带系数。对低频子带,将原图的低频子带系数直方图匹配到直方图均衡化后图像的对应系数直方图上。对各个高频子带,则先进行阈值去噪,再将原图的各个高频子带系数直方图匹配到直方图均衡化后图像的对应系数直方图上。最后,经NSCT重构得到增强后的最终图像。实验结果表明,本文方法增强效果明显优于直方图均衡化,与Contourlet变换增强法相比,实验所采用的两组图像的图像评价函数(EMEE)值分别提高了24.05%、16.97%、13.29%和20.63%,且与NSCT域非自适应增强法(人工选取参数)的处理效果相当。该方法无需手工调节参数,具有自适应性和实用性强的优点。
Abstract
As the image enhancement algorithm of NonSubsampled Contourlet Transform(NSCT) domain has to adjust its parameters manually and can not enhance images adaptively, this paper proposes an adaptive image enhancement algorithm by combining histogram equalization with NSCT domain enhancement. The algorithm firstly performs the histogram equalization to the original low-contrast and noisy image. Then, it conducts the NSCT decomposition on the original image and the histogram equalized image to obtain the low frequency subband coefficients and a series of the high frequency directional subband coefficients. In the low frequency subband, the transform coefficient histogram of the original image is mapped to that of the equalized image. In each high frequency subband, the transform coefficient histogram of the original image is mapped to that of the equalized image after threshold denoising. Finally, the enhanced image is obtained by reconstruction of the modified NSCT coefficients. Experimental results show that the enhancement of the proposed algorithm is superior to that of classical histogram equalization method. As contrasted with Contourlet transform enhancement in two group of images, its evuluation function EMEE(Measurement of Enhanement by Entropy) values increase by 24.05%, 16.97%, 13.29% and 20.63% , respectively, which corresponds to that of NSCT non-adaptive enhancement(selecting optimal parameters manually) well. Moreover, this algorithm does not need manual adjusting parameters, and is characterized by good adaptability and practicability.
周妍, 李庆武, 霍冠英. 基于非下采样Contourlet变换系数直方图匹配的自适应图像增强[J]. 光学 精密工程, 2014, 22(8): 2214. ZHOU Yan, LI Qing-wu, HUO Guan-ying. Adaptive image enhancement based on NSCT coefficient histogram matching[J]. Optics and Precision Engineering, 2014, 22(8): 2214.