Perceptual Hashing for Large-Scale Multimedia Search

Li Weng, I. Hong Jhuo, Wen Huang Cheng

Research output: Chapter in Book/Report/Conference proceedingChapterpeer-review

2 Scopus citations

Abstract

This chapter presents perceptual hashing technique together with a particular category of algorithms called perceptual hash algorithms. These algorithms are used for generating hash values from large-scale multimedia objects, such as images, audio, and video. The chapter focuses on unsupervised perceptual hash algorithms and supervised perceptual hash algorithms. Perceptual hashing is one of the approaches that seek compact representations of multimedia data. Perceptual hashing mainly consists of two parts: hash generation and hash verification. Hash generation is the focus of hash algorithm design. There are a few essential components: feature extraction, feature transformation, dimension reduction, quantization, and randomization. Hash verification is typically made simple in order to be fast. The basic properties of perceptual hashing are robustness and discrimination. Kernelized locality sensitive hashing is an extension of locality sensitive hashing. Semi-supervised hashing is a hash algorithm that takes both semantic relevance and “maximal bit variance” into account.

Original languageEnglish
Title of host publicationBig Data Analytics for Large-Scale Multimedia Search
Publisherwiley
Pages239-265
Number of pages27
ISBN (Electronic)9781119376996
ISBN (Print)9781119376972
DOIs
StatePublished - 1 Jan 2019

Keywords

  • Discrimination
  • Hash generation
  • Hash verification
  • Kernelized locality sensitive hashing
  • Large-scale multimedia objects
  • Perceptual hashing
  • Robustness
  • Supervised perceptual hash algorithms
  • Unsupervised perceptual hash algorithms

Fingerprint

Dive into the research topics of 'Perceptual Hashing for Large-Scale Multimedia Search'. Together they form a unique fingerprint.

Cite this