对数域中自动色彩均衡化快速算法  

The Fast Automatic Color Equalization Algorithm in Logarithm Domain

作　　者：肖志云[1] 李新科[1] 轩波[2,3] 胡晰远[2] 

机构地区：[1]内蒙古工业大学电力学院,呼和浩特010080 [2]中国科学院自动化研究所高技术创新中心,北京100190 [3]北京多维视通技术有限公司,北京100190

出　　处：《计算机辅助设计与图形学学报》2013年第3期390-396,共7页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金青年基金(61101219);内蒙古自治区高等学校科学研究项目(NJ10074)

摘　　要：对于存在颜色一致区域或低对比度区域的图像,在校正后往往会出现"灰化"现象.尽管自动色彩均衡化(ACE)算法采用整幅图像作为计算子集进行颜色校正,在一定程度上克服了"灰化"现象,但导致了计算量的大幅增加.针对这一问题,提出一种在对数域内融合了局部特征和全局特征的克服"灰化"现象的算法.首先将原图像变换到对数域,实现光照与反射系数的解耦;然后对变换后的图像进行局部特征计算/侧抑制调整,同时计算图像的全局特征;最后利用标准的增益与偏移方法将两者的融合图像映射到正常显示的区域,并改善图像的对比度.实验结果表明,该算法不仅可以大大减少计算量并有效克服"灰化"现象,而且可以获得满意的主观视觉效果.Images containing constant color areas or low contrast areas usually suffer from Gray-out phenomenon after correction. By using the whole region of image as the computational subset, Automatic Color Equalization （ACE） can reduce Gray-out phenomenon. However, it increases computational cost greatly. Aim at this problem, a fast algorithm which fuses local and global characteristic in logarithm domain is proposed. First, the coupling between illumination and reflectance coefficient is decoupled by logarithm transformation. Then, the local characteristic and lateral inhibition of the transformed image are computed and adjusted. The global characteristic is then computed. Finally, the fused image of the last two steps is mapped to a suitable domain using canonical gain/offset method so that the final image can display correctly with improved contrast. Experimental results show that this improved algorithm not only can reduce the computational cost and reduce the gray-out phenomenon effectively, but also can provide better color correction performance.

关 键 词：自动色彩均衡化算法 侧抑制机制 颜色校正 对数变换 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]