改进的Gray World-Retinex图像自动白平衡方法  

Modified Automatic White Balance Method for Gray World-Retinex Image

作　　者：张钰[1] 姚素英[1] 张娜[1] 金则群 

机构地区：[1]天津大学电子信息工程学院,天津300072 [2]士兰微电子股份有限公司,杭州310012

出　　处：《数据采集与处理》2008年第6期673-677,共5页Journal of Data Acquisition and Processing

基　　金：国家自然科学基金(50323005)资助项目

摘　　要：提出了一种改进的白平衡方法,改善了Gray World-Retinex算法对色彩丰富度低、近白(灰)点数量较少的图像白平衡后仍然偏色的缺点。使用图像Cb和Cr分量的样本标准差分析图像色彩丰富度,并改进了确定图像近白(灰)点数量的方法,最终定量地给出适合采用Gray World-Retinex白平衡法的图像特征,对于样本标准差极低而且近白(灰)点很少的图像,给出一种求极大值增益的方法进行白平衡。对40幅色彩丰富度和近白(灰)点数量不同的偏色图像进行验证,并与Gray World-Retinex算法作了对比。结果表明,本文算法对于具有不同色彩丰富程度和近白(灰)点数量的偏色图像,都可以获得较理想的白平衡效果。该算法复杂度低,对图像实时处理具有很大吸引力。A modified automatic white balance method is proposed. The method overcomes the defects of Gray World-Retinex algorithm for images with small sample standard deviation value and little number of near white points. By calculating the sample standard deviation values of Cr and Cb, and ascertaining the number of near white points, the degree of color sufficiency on one image, the characteristics of those images for the suitable Gray World-Retinex algorithm are determined. The maximum gain method is used to accomplish white balance for those images with small sample standard deviation value and little number of near white points. Experimental results show that the algorithm is not limited by the degree of color sufficiency and the number of near white points on the images.

关 键 词：自动白平衡 样本标准差 GRAY World—Retinex算法 CMOS图像传感器 

分 类 号：TN911.73[电子电信—通信与信息系统]