焦点引导的带颜色恢复的多尺度Retinex算法  

Focus-guided multi-scale Retinex with color restore algorithm

作　　者：窦易文[1,2] 周鸣争[1] 唐肝翌[1] 卢桂馥[1] 

机构地区：[1]安徽工程大学计算机科学与信息学院,安徽芜湖241000 [2]东华大学信息科学与技术学院,上海201620

出　　处：《计算机工程与应用》2013年第2期207-210,270,共5页Computer Engineering and Applications

基　　金：江苏省2010年度普通高校研究生科研创新计划项目(No.178);安徽省自然科学基金(No.KJ2007B056)

摘　　要：针对低照度下高动态范围图像在应用多尺度Retinex算法进行图像增强时,易出现光晕现象和颜色失真等问题,提出了一种可以由焦点引导的带颜色恢复的多尺度Retinex(FgMSRCR)算法。通过自底向上视觉注意模型对低照度下高动态范围图像进行焦点提取,保留于相异焦点集中,对整幅图像应用带色彩恢复的多尺度Retinex算法(MSRCR)进行图像增强,再对焦点区域进行保持亮度和颜色比例的增强,利用高斯平滑滤波器完成图像增强任务。实验结果表明,与常见的图像增强算法和其他Retinex算法相比,该算法能够在保持颜色恒常的前提下有效地克服光晕现象。A focus-guided multi-scale Retinex with color restore algorithm is presented to solve artificial halos and color distor- tion after applying multi-scaled Retinex algorithm to enhance high dynamic range images under low illumination. In initial stage, using bottom-up visual saliency model to extract saliency focus, a dissimilar focus sets can be obtained. Then utilizing multi-scaled Retinex algorithm the whole image is enhanced and the focus area in enhanced image is second enhanced under color and bright- ness rates preservation. Furthermore, applying Gaussian filter to smooth the enhanced image gets the final image. Experimental results show that the presented algorithm can overcome artificial halos and keep color constant comparing with general enhance algorithms and other Retinex algorithms.

关 键 词：图像增强 颜色恒常性 RETINEX算法 视觉注意模型 

分 类 号：TP37[自动化与计算机技术—计算机系统结构]