Abstract: Image retouching is a technique widespreadly used in image tampering.To implement image retouching blind detection, a blind identification algorithm for image was proposed. The algorithm firstly searches for each block in image, and inserts each block into the KD tree. Thus obtain the same value or the nearest match block, then it uses the hierarchical clustering of the position vector to eliminate the scattered matching. Finally, it applies the 7-tap Laplacian filter to count weight of zero connectivity of suspicious blocks to eliminate false positives, thereby locate accurately the tampered area. Experiments show that this method can efficiently and accurately identify the use of a class of image manipulation techniques for uncompressed images and high quality compressed images. When it is applied to the images that have higher compression levels, accurate results are also obtained if the region that has been retouched is sufficiently large.

Key words: healing brush, image retouching, blind identification, KD tree

摘要： 图像润饰是一种广泛应用的图像篡改手段。为了对润饰的图像实施盲检测，提出了一种图像盲鉴别算法。该算法首先查找图像中每个分块并把其插入到KD树中，搜索到值相同或最近的粗略匹配块，然后使用位置向量的分层聚类法群集块对消除杂散配对，最后应用7-tap拉普拉斯过滤器并统计可疑块的零连通分量来消除误报，从而定位出精确的润饰篡改区域。实验表明,该方法能有效精确地识别出修复刷对非压缩图像和高品质压缩图像等一类图像的润饰篡改技术的使用。当应用到压缩级别较高的图像时，如果润饰的区域足够大，依然会获得准确的结果。

关键词: 修复刷, 图像润饰, 盲鉴别, KD树