关键词: 高光去除, 最大漫反射色度, 线性模型

Abstract: Highlight removal from a single image is generally much challenging.Summarized most single-image-based methods，usually require color segmentation or user assistance for highlight detection.Based on statistical law of color highlight images，this paper finds a key observation that the maximum diffuse chromaticity in local patches in colorful images generally changes smoothly.Using this property，it then estimates the maximum diffuse chromaticity values of the specular pixels by directly applying linear model to the maximum fraction of the color components of the original image，such that the maximum diffuse chromaticity values can be propagated from the diffuse pixels to the specular pixels.The diffuse color at each pixel can finally be computed.Compared to traditional methods，this approach can get much better results after removing highlights.Furthermore，it is well suited for real-time applications，due to its simplicity and parallelism.

Key words: highlight removal, maximum diffuse chromaticity, linear model