珞珈三号01星冰雪/云场景影像偏色校正处理

Color Correction of Luojia3-01 Satellite Images with Partial Snow or Cloud Cover

珞珈三号01星采用Bayer成像模式,其获取的绿色通道信号相对较强,但由于相机通常采用固定的曝光与增益设置,导致珞珈三号01星拍摄的含冰雪/云覆盖影像会出现严重偏绿现象。针对该问题,提出了一种结合颜色通道补偿与直方图间互相关关系的直方图重叠白平衡方法。为提升影像的细节信息,通过低灰度值区域的分布比例构建伽马函数,实现拉伸处理;为提升各颜色通道分布的相似度,使用影像的标准差信息计算权重,补偿衰减通道,并基于YCbCr颜色空间保持亮度信息,避免影像背景亮度变化;为改善偏色情况并保持色度均值的重合,通过互相关关系进行色度直方图重叠处理,根据冰雪/云场景影像的特征自适应调整参数。对校正后的结果影像进行主观分析与客观评价,并与多种主流的颜色校正方法进行对比验证。实验结果表明,所提方法颜色校正效果更好、影像地物颜色更自然,且能更好地保留地物信息。

Objectives The green channel received by Luojia3-01 satellite in Bayer imaging mode has relatively strong signals.The camera often adopts a fixed exposure and gain setting for images containing high saturation pixel values,which comprehensively leads to the color deviation of the images with partial snow or cloud cover captured by Luojia3-01 satellite.Methods To solve the above problems,a histogram overlapping white balance method combining color compensation and cross-correlation is proposed.First,in order to further improve the details of the image,the Gamma function is constructed through the distribution proportion of the region of low gray value to enable adaptive gamma correction processing.In order to improve the similarity of the distribution of color channels,the standard deviation of the image is used to calculate the weight to compensate for the attenuation channels,and the brightness information is maintained based on YCbCr color space to avoid the background brightness change of the image.Then,in order to improve the color deviation and maintain the overlap of the color mean,the color histogram is overlapped by using the cross-correlation relationship,and the parameters are adjusted according to the features of the images with partial snow or cloud cover.Finally,subjective analysis and objective evaluation of the corrected images are carried out.Results The proposed method has the best processing effect in retaining the information of other ground objects except snow or cloud,and the information entropy value is the largest.The calculated value of the color correlation index obtained by processing the scene images with different amounts of snow and ice or cloud coverage also reaches the optimal value,and the color correlation is the highest compared with other methods.Other color correction methods,such as gray-world algorithm and gray-edge algorithm,are not effective,and it is difficult for other methods to achieve a balance between improving the color cast and preserving image information.Conclusions The proposed method has better color correction effect,more natural color,and can better preserve feature information.