光学与光电技术, 2018, 16 (6): 25, 网络出版: 2018-12-17  

基于Wasserstein直方图欧氏度量的图像重复区域盲检测算法

A Blind Algorithm for Duplication Region Detection Based on Wasserstein Histogram Euclidean Metric
作者单位
南京工业大学计算机科学与技术学院, 江苏 南京 211816
摘要
提出了一种利用Wasserstein重心坐标作为图像描述子的图像重复区域盲检测算法。通过度量各图像块Wasserstein重心坐标间的最小欧式距离确定图像重复区域初始搜索块,再对初始搜索块进行扩展, 利用改进的PatchMatch算法寻找相关度高的图像块, 从而搜索出完整的图像重复区域。实验结果表明, 提出的基于Wasserstein直方图欧氏度量的图像重复区域盲检测算法能够检测出复制粘贴的伪造区域, 对于带有旋转和缩放等几何攻击的伪造区域检测率维持在90%以上。该算法对复制粘贴、尺度变化和旋转几何攻击具有良好的鲁棒性。
Abstract
By using Wasserstein's concept of the barycentric and barycentric coordinates, a blind region detection algorithm for image region-duplication forgery using Wasserstein barycentric coordinates as image descriptors is proposed. The initial search block of the image is determined by measuring the minimum Euclidean distance between the Wasserstein barycentric coordinates of each image block. The initial search block is extended, and a patch with high correlation is searched for by using the improved PatchMatch algorithm to search for a complete image repetition area. Experimental results show that the proposed algorithm can detect the repeated region by using blind detection algorithm based on the Wasserstein histogram Euclidean metric. The copying and pasting forged regions are also very robust for forged regions with rotation and zoom geometry attacks.

杜振龙, 叶超, 李晓丽, 沈海洋, 宋国美. 基于Wasserstein直方图欧氏度量的图像重复区域盲检测算法[J]. 光学与光电技术, 2018, 16(6): 25. DU Zhen-long, YE Chao, LI Xiao-li, SHEN Hai-yang, SONG Guo-mei. A Blind Algorithm for Duplication Region Detection Based on Wasserstein Histogram Euclidean Metric[J]. OPTICS & OPTOELECTRONIC TECHNOLOGY, 2018, 16(6): 25.

关于本站 Cookie 的使用提示

中国光学期刊网使用基于 cookie 的技术来更好地为您提供各项服务,点击此处了解我们的隐私策略。 如您需继续使用本网站,请您授权我们使用本地 cookie 来保存部分信息。
全站搜索
您最值得信赖的光电行业旗舰网络服务平台!