基于HSV和CLAHE的复制粘贴篡改检测算法

发布时间：2018-06-26 10:17

本文选题：色调、饱和度、明度颜色空间 + 限制对比度自适应直方图均衡化　；参考：《北京邮电大学学报》2016年06期

【摘要】：针对现有的局部特征提取算法,如尺度不变特征变换、加速稳健特征等对面积较小的篡改区域和平滑区域无法准确提取特征点进行复制粘贴篡改检测的问题,提出了一种基于色调、饱和度、明度(HSV)颜色空间和限制对比度自适应直方图均衡化(CLAHE)的2阶段篡改检测方法.第1阶段,将待检测图像从红、绿、蓝颜色空间转换到(HSV)颜色空间进行加速稳健特征提取;第2阶段,针对平滑区域的篡改,通过CLAHE算法对图像进行特征增强之后,再采用加速稳健特征进行特征提取;然后利用广义2近邻算法进行特征匹配,并利用随机抽样一致性算法剔除错误的匹配点;最后通过形态学操作实现篡改区域的标示定位.实验结果表明,该算法能抵抗小面积区域和具有非显著视觉特征的平滑区域复制粘贴篡改,并对旋转、缩放等后处理攻击具有一定的鲁棒性.
[Abstract]:The existing local feature extraction algorithms, such as scaling invariant feature transformation, accelerating robust feature and so on, can not accurately extract feature points from small tamper area and smooth region for copying and paste tampering detection. A two-stage tampering detection method based on hue, saturation, HSV color space and restricted contrast adaptive histogram equalization (CLAHE) is proposed. In the first stage, the image to be detected is transformed from red, green and blue color space to (HSV) color space for fast robust feature extraction. In the second stage, for the tampering of smooth region, the image is enhanced by CLAHE algorithm. Then we use accelerated robust feature extraction; then we use the generalized 2 nearest neighbor algorithm to match the feature, and use random sampling consistency algorithm to eliminate the wrong matching points; finally, we use morphological operations to implement the tampered region marking location. The experimental results show that the proposed algorithm can resist small areas and smooth areas with non-significant visual features, and is robust to post-processing attacks such as rotation, scaling and so on.
【作者单位】：中国农业大学理学院;邢台学院数学与信息技术学院;北京邮电大学智能通信软件与多媒体北京市重点实验室;
【基金】：国家自然科学基金项目(61370195,U1536121)
【分类号】：TP391.41

【相似文献】