Ring Partition and DWT Based Perceptual Image Hashing with Application to Indexing and Retrieval of Near-Identical Images

Abstract

The perceptual image hashing function maps an input image to a fixed size short binary string called the perceptual hash. The perceptual image hashing has been used for image authentication, image tamper detection and digital watermarking applications. In this paper, the perceptual hash value is generated using ring partition and Discrete Wavelet Transform (DWT). An rotation-invariant secondary image is constructed using ring-partition technique. Then, the DWT is applied on the secondary image using the various mother wavelets, to generate a final hash value. The perceptual hash was used to index and retrieve near-identical images from the database. The precision-recall curves are used to evaluate the performance of the application. The experiments shows that the proposed hashing algorithm is robust against content-preserving operations like image rotation, image scaling, Gaussian low-pass filtering, JPEG compression, brightness adjustment, and gamma correction.