Video copy detection method based on multi-feature Hash

Abstract

The invention discloses a video copy detection method based on multi-feature Hash, which mainly solves the problem that detection efficiency and detection accuracy cannot be effectively balanced in the exiting video copy detection algorithm. The video copy detection method based on multi-feature Hash comprises the following realization steps of: (1) extracting the pyramid histogram of oriented gradients (PHOG) of a key frame as the global feature of the key frame; (2) extracting a weighted contrast histogram based on scale invariant feature transform (SIFT) of the key frame as the local feature of the key frame; (3) establishing a target function by a similarity-preserving multi-feature Hash learning SPM2H algorithm, and obtaining L Hash functions by optimization solution; (4) mapping the key frame of a database video and the key frame of an inquired video into an L-dimensional Hash code by virtue of the L Hash functions; (5) judging whether the inquired video is the copied video or not through feature matching. The video copy detection method based on multi-feature Hash disclosed by the invention is good in robustness for multiple attacks, and capable of being used for copyright protection, copy control and data mining for digital videos on the Internet.

Description

technical field [0001] The invention belongs to the technical field of information security, specifically a method for video feature extraction and copy detection, which can effectively resist conventional video attacks, geometric attacks and combined attacks, and can be used for copyright protection of digital videos on the Internet, video content Analysis and copy control areas. Background technique [0002] With the continuous advancement of digital technology and the increasing popularity of computer networks, multimedia data is gradually becoming an important source of information for people, especially more and more Internet video sites. Digitized multimedia data is easy to obtain, easy to copy and spread quickly, which not only provides great convenience for the access of multimedia information, but also greatly improves the efficiency and accuracy of information expression, but the piracy problems and copyright disputes caused by this And data management has also be...

AI Technical Summary

Problems solved by technology

First of all, most of the existing methods use a single feature to describe the video content, but different visual features have completely different description functions. For example, the global feature can effectively describe the global information of the video, but the robustness to global attacks is weak. Local features can effectively describe the local information of the video, but they are sensitive to local attacks of the video. Therefore, the video information constructed by a single feature cannot fully describe the video content and lacks extensive robustness. However, directly cascading multiple visual features will Cause dimensionality disaster and reduce detection efficiency; secondly, most of the existing indexing methods aim to improve detection efficiency while ignoring detection accuracy

Images