摘要
描述了光照方向的重要意义及光线矛盾对鉴别图像真伪的作用,总结和分析了现有的针对单幅数字图像点光源情况下的光照方向估计的典型方法,着重对Pentland,Lee,Zheng,Yang算法和Farid算法进行深入研究,阐述了在算法实现过程中遇到的各种问题和应对手段,借助模拟生成图像、实际拍摄图像和数码合成图片的验证结果对5种不同算法进行了性能比较,并拓展应用到通过判断数字图像中不同对象的光照方向是否一致对图像进行篡改检测,最后探讨了估计数字图像光照方向存在的问题及未来研究方向。
In this paper, the significance of illumination direction and the role in the image forgery by detecting lighting inconsistencies are described. A summary and a detail analysis of existing typical methods for estimating illuminant direction from only a single digital image in the case of point light source are given, especially the Pentland, Lee, Zheng, weighted average and Farid algorithm are studied. The various problems encountered and the coping methods in algorithm implementation are discussed. As well as the performance comparison of each algorithm by the experiment result to synthetically generated images, real images and forgery images. At the same time, the application to identify the fogery image by detecting lighting inconsistencies is expanded. Finally, the discussion of unresolved problems and future directions about the illuminant direction estimation are presented.
出处
《电视技术》
北大核心
2011年第15期41-45,共5页
Video Engineering
基金
国家自然科学基金项目(60872116)
关键词
光照方向
点光源
朗伯表面
光源偏角
光源倾角
图像取证
illumination direction
point light source
lambert surface
tilt angle
slant angle
image forensics
