计算机工程与应用 ›› 2014, Vol. 50 ›› Issue (10): 176-179.

• 图形图像处理 • 上一篇    下一篇

基于快速双边滤波的图像高光去除研究

卢桂荣1,汤景凡2,姜  明2   

  1. 1.杭州科技职业技术学院 信息工程学院,杭州 311402
    2.杭州电子科技大学 计算机学院,杭州 310018
  • 出版日期:2014-05-15 发布日期:2014-05-14

Research on image highlight removal based on fast bilateral filter

LU Guirong1, TANG Jingfan2, JIANG Ming2   

  1. 1.College of Information and Engineering, Hangzhou Polytechnic, Hangzhou 311402, China
    2.College of Computer, Hangzhou Dianzi University, Hangzhou 310018, China
  • Online:2014-05-15 Published:2014-05-14

摘要: 在各向异性的物体中,高光被视为是漫反射分量以及镜面反射分量的一种线性组合。单幅图像的高光去除是计算机视觉中一项非常有挑战性的课题。很多方法试图将漫反射分量、镜面反射分量进行分离,然而这些方法往往需要图像分割等预处理过程,方法鲁棒性较差且比较耗时。基于双边滤波器设计了一种高效的高光消除方法,该方法利用最大漫反射色度存在着局部平滑这一性质,使用双边滤波器对色度的最大取值进行传播与扩散,从而完成整幅图像高光去除。方法采用一种加速策略对双边滤波器进行速度优化,与目前流行的方法相比,有效提升了方法的执行效率。与传统方法相比,该方法高光去除效果更好,处理速度更快,非常适用于一些实时应用的场合。

关键词: 快速高光去除, 双边滤波, 最大漫反射色度, 双色反射模型, 线性插值, 图像处理

Abstract: In inhomogeneous objects, highlights are linear combinations of diffuse and specular reflection components, and highlight removal from a single input image is treated as a non-trivial task in computer vision. A number of methods have been proposed to separate or decompose diffuse and specular reflection components. However, all methods require explicit color segmentation process, causing low robustness and high computational complexity. In this paper, an effective specula highlight removal method using the bilateral filter is proposed. The method utilizes a key property-the maximum diffuse chromaticity in local patches in colorful images generally changes smoothly. Using such observation, then the maximum diffuse chromaticity values of the specular pixels are computed by directly applying bilateral filter to the maximum fraction of the color components of the original image, such that specular highlight can be removed ultimately. The fast bilateral filtering techniques is utilized to enable the method to run much faster than the state-of-the-art. Compared to traditional methods, the method can get much better results and faster speed, which is very well suited for real-time applications.

Key words: fast highlight removal, bilateral filter, maximum diffuse chromaticity;dichromatic reflection model, linear interpolation, image processing