Abstract: To solve the problem of texture information loss due to saturation and highlight of specular surface reflection, a method of removing highlight based on multi view image feature（MSF） matching is proposed. Firstly, two images from different viewpoint are selected to be reference image and auxiliary image. They are converted to HSV space to adjust the brightness. Next, the homography matrix is estimated from the matched feature points extracted from the two images. Then the homography is utilized to transform the auxiliary image to the same viewpoint of the reference image. Finally, the MSF algorithm is used to detect the highlight from the reference image and the highlight region is employed as a mask in auxiliary image to obtain the texture that can be used to fill the missing texture in the reference image. The experimental results show that proposed method is suitable for various real scene highlight removal. It can maximally recover the texture details of the highlight region at the same time to guarantee the peak signal-to-noise ratio and the structure similarity of the image. The visual effect after highlight removal is desirable.

Key words: multi view images, brightness adjustment, homography matrix, high pixel detection, high light removal

摘要： 针对物体镜面反射使拍摄图像出现饱和高光导致的纹理信息丢失问题，提出了一种基于多视角图像特征匹配的高光去除方法。首先，选取两幅不同视角的图像作为基准图和辅助图，转换到HSV空间进行亮度调整，将亮度统一后的两幅图像进行特征点匹配，估算出单应性变换矩阵，然后利用该矩阵将辅助图透视变换到与基准图同一视角，最后利用MSF算法对基准图进行高光检测，将检测出来的高光区域遮罩在辅助图的相应位置选择纹理信息对基准图进行填补并修复缺失纹理，从而去除基准图的高光。实验结果表明，所提出的方法适用于多种真实场景图像的高光去除，且在保证图像峰值信噪比和结构相似性有所提高的同时能最大限度地恢复高光区域的纹理细节信息，高光去除后的视觉效果更好。

关键词: 多视角图像, 亮度调整, 单应性变换矩阵, 高光像素检测, 高光去除