Abstract:Perceptual image hashing methods are often applied in various objectives, such as image retrieval, finding duplicate or near-duplicate images, and finding similar images from large-scale image content. The main challenge in image hashing techniques is robust feature extraction, which generates the same or similar hashes in images that are visually identical. In this article, we present a short review of the state-of-the-art traditional perceptual hashing and deep learning-based perceptual hashing methods, identifying the best approaches.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is how to effectively detect, prevent, and respond to complex cybercrimes in the field of multimedia information security, especially crimes involving forged images or the spread of illegal content. Specifically, the paper focuses on the applications of perceptual hashing techniques in tasks such as image retrieval, image authentication, digital watermarking, image copy detection, tampering detection, image indexing, multimedia forensics, and no - reference image quality assessment. The main challenge of these techniques lies in robust feature extraction, that is, the method of generating the same or similar hash values in visually identical images.
The paper specifically mentions that with the popularization of the Internet and smart devices, a large number of images and videos are captured, stored, and shared, which has led to an increasingly serious problem of unauthorized distribution of multimedia files and illegal access. In order to protect the security of multimedia content, many researchers have proposed similarity - based image detection or retrieval methods using perceptual hash functions. These methods can extract specific features from images and generate 64 - bit or 128 - bit numerical values, called hash codes, which are used to form "perceptually equivalent" image content.
In addition, the paper also explores the development of deep hash techniques in recent years, which is a method of combining perceptual hashing with deep - learning techniques, mainly used for retrieving or detecting similar images from large - scale datasets. By comparing traditional perceptual hashing methods and deep - learning - based hashing methods, the paper aims to identify the best perceptual hashing algorithm and evaluate its robustness under different attack conditions. For example, the paper uses the USC - SIPI dataset to evaluate four traditional perceptual hashing methods (Ring Partition and Invariant Vector Distance (RP - IVD), Selective Sampling for Salient Structure Features (SS - Salient - SF), pHash, and F - DNS), and tests their performance under operations such as brightness adjustment, gamma correction, salt - and - pepper noise, multiplicative noise, 3×3 Gaussian low - pass filtering, JPEG compression, scaling, rotation, and watermark embedding.
In summary, the goal of the paper is to review the current state - of - the - art traditional perceptual hashing techniques and deep - learning - based perceptual hashing techniques, and experimentally evaluate the performance of these techniques in practical applications, with the expectation of providing effective solutions for the authentication and protection of multimedia content.