中文摘要：

反射分量分离是计算机视觉和数字图像处理中的一个重要问题。尽管已有很多基于单张图像的反射分量分离方法,但这些方法只能分离图像中彩色区域的反射分量并会在等色区域产生严重的噪声。本文提出一种能够分离图像彩色和等色区域反射分量的方法。彩色和等色区域高光的共同特征是亮度在局部区域中逐渐变化,因此本文算法首先将亮度信息融入传统的无高光图像,提出能够区分不同亮度等色区域的改进的无高光图像;然后提取局部位置空间亮度差异特征和局部颜色空间亮度差异特征,并用K-Means方法检测图像中的镜面反射像素;最后用颜色传递方法估计出漫反射分量,实现漫反射和镜面反射分量的分离。实验结果表明本文算法能够同时有效地分离彩色和等色区域的反射分量。本文算法扩展了反射分量分离方法的应用范围。

英文摘要：

Reflection components separation is an important problem in computer vision and image processing. Though several single image based reflection components separation methods have been developed for chromatic surfaces, they are challenging for achromatic regions, where heavy noise is caused. To address the problem, a new reflection components separation method for both chromatic and achromatic surface is proposed. Based on the observation that specular reflection causes local intensity variation in both chromatic and achromatic regions, an adaptive specular-free image was first introduced by incorporating intensity into a traditional specular-free image, which can distinguish achromatic regions with different intensity; local spatial intensity difference feature and local color space intensity difference feature were then extracted and K-Means method was applied to detect specular pixels; finally, a color transfer technique was used to estimate the diffuse component. Experimental results show that the proposed method separates reflections on both chromatic and achromatic surfaces effectively. The proposed method extends the range of application of reflection components separation methods.