Abstract: Based on the assumption of spatially smooth illuminant, traditional Retinex enhancement methods suffer the existence of halo artifacts. According to its drawback, add to the advantage of MSRCR, a fuzzy multiple-scale Retinex color image enhancement method is proposed. The new Retinex method restricts the center/surround spatially contrast operation in region with homogeneous illumination. Experimental results demonstrate that the method efficiently reduces halo artifacts. Meanwhile, it uses an adaptive Gaussian mask to reduce the computational time. Experimental results are given to show its effectiveness.

Key words: fuzzy multiple-scale retinex, color image enhancement, MSRCR algorithm

摘要： 由于中心环绕Retinex算法中假设场景中光照是平缓变化的，所以在图像明暗对比强烈处易出现光晕现象。针对Retinex传统算法的固有缺陷，结合MSRCR算法在色彩恢复上的优势，提出了一种模糊多尺度Retinex彩色图像增强方法（FMSRCR）。FMSRCR使中心环绕空间对比运算仅在光照强度相近的区域中进行，克服了光照不均的影响。同时采用自适应高斯模，减少了卷积运算量。通过实验证明该方法是有效的。

关键词: 模糊多尺度Retinex, 彩色图像增强, MSRCR算法