基于NLEMD的Retinex图像增强算法  

Retinex image enhancement algorithm based on NLEMD

作　　者：蒋永馨[1,2] 王孝通[3,2] 徐晓刚[3,4] 黄华[3,2] 

机构地区：[1]海军大连舰艇学院基础部,大连116018 [2]海军大连舰艇学院光电技术研究所,大连116018 [3]海军大连舰艇学院航海系,大连116018 [4]海军大连舰艇学院装备系统与自动化系,大连116018

出　　处：《计算机应用》2008年第S2期161-163,170,共4页journal of Computer Applications

基　　金：国家自然科学基金资助项目(60473141);辽宁省自然科学基金资助项目(20082176);浙江大学CAD&CG国家重点实验室开放基金资助项目

摘　　要：为了将已有的受光照影响严重的图片恢复出细节信息更加丰富且光照平衡的图片,提出了基于NLEMD的Retinex图像增强算法。该算法采用多幅图像(其中包括一副待增强的图像和将待增强图像分别提高n1,n2,…,nm倍照射系数的多幅图像)协同NLEMD和Retinex算法来实现图像的增强。采用将待增强图像提高n倍照射系数来获得多幅图像的目的是为了获得更丰富的图像暗处或明处的细节信息,然后利用NLEMD充分提取这些已知图像的各个频率信息;选取其中一幅图像经过NLEMD分解后所得的剩余分量,用Retinex增强算法增强;最后叠加所有图像的高频信息和Retinex算法增强后的剩余分量,得到增强后的图像。实验结果表明,算法增强后的图像克服了传统Retinex算法增强后图像对比度低和引入了噪声的缺陷,且暗处细节信息更加丰富,获得了更舒适的视觉效果。The Retinex image enhancement algorithm based on Neighborhood Limited Empirical Mode Decomposition(NLEMD) was proposed.An image is enhanced by using serial images through NLEMD and Retinex to derive more information.Then,it picks up all the frequency information of all images with NLEMD.A remain weight of a selected image is enhanced with Retinex.At last,it pluses high frequency information of all images and the enhanced remain weight to enhance the images.The experimental result indicates that the proposed algorithm overcomes the disadvantages of traditional Retinex algorithm that the contrast of the enhanced image is lower and the noise of the enhanced image is introduced.And it has more detail in dark.The result is more comfortable to people's eyes.

关 键 词：图像增强 经验模式分解 RETINEX 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]