基于保边滤波器的单幅图像高光去除

Highlight Removal from Single Image Using Edge-Preserving Filter

研究图像分割优化问题,由于图像中存在高光中,造成干扰,提出一种简单但有效的单幅图片高光去除方法。传统单幅图像高光去除方法,一般需要进行图像分割预处理,或者进行交互输入。提出的新方法无需图像分割预处理,也不要交互输入,而是使用漫反射色度估计值作为引导值,对最大色度图应用边界保留滤波器,并最终得到去高光后的图片。实验结果比较表明,与传统方法比较,该方法不但能够去除单色物体表面高光,对于复杂纹理表面高光非常有效,且输出图像在非高光区更接近输入图像,提高了图像分割效果。

This paper introduced a simple but effective method to remove highlight from a single given image. Tra- ditionally, image segmentation of preproeessing or interaction between human and computer is necessary for the elimi- nation of the highlight, however, it is not necessary to use our method. In this paper, a method based on the law that maximum diffuse chromaticity in local patch in any colorful image changes smoothly was introduced. To remove the highlight from the given image, an edge-reserve filter was applied to smooth the maximum chromaticity, with the approximation of maximum diffuse chromaticity applied as the smoothing guidance. Compared with a conventional method, it shows a better performance in removing the highlights existed in single-color images as well as heavy-tex- tured images, and the output image of the introduced method is closer to the input image in diffuse area.