基于图像正则化和DCT变换的感知哈希技术

Perceptual Hashing Algorithm Based on Image Normalization and DCT Transform

图像感知哈希技术是一门较新型的技术,哈希提取过程的关键步骤是特征提取,传统的基于DCT变换的感知哈希技术抗几何攻击能力较差。试图在图像生成哈希之前,将几何形变对图像的影响去除,首先将图像进行正则化,使图像具有几何不变性,然后再进行DCT特征系数提取,最后通过量化、编码生成最终哈希。该算法可以抵抗任意的仿射变换。

The technique of perceptual hashing is a new technology relatively,and the feature extraction is the most important step in the extraction process,the traditional and DCT-based hashing technique is less able to resist the geometric attacks. This paper attempts to remove the influence of geometric effects on image before generating the image hash.First,normal- ize the image to make it invariant to any affine transforms,then extract the feature coefficient through DCT transformation, and finally generate the final hash through quantization and encoding.The algorithm can resist any affine transformations.