Method of proceeding video frequency searching through video frequency segment

Abstract

The invention belongs to video search technique area especially, related to the new method for searching video based on a segment of video. Aim at disadvantages of current technique such as low searches precision and slow searches speed, the invention puts forward and solves two difficult problems under frame of optimal matching in graph theory: automatic obtaining similar segments and evaluating degree of similarity between two segments. Model of weighted bipartite graph is built. Examining continuity of similar frames obtains similar segments in preliminary. Maximum weighted matching between these segments and queried segment is calculated out by using Kuhn_Munkres method. The match determines whether two segments are similar or not as well as boundary of similar segments. Maximum weight shows degree of similarity between two segments. Comparing with prior art, experimental result indicates the invention obtains higher precision ratio, recall ratio and quicker searching speed.

Description

technical field [0001] The invention belongs to the technical field of video retrieval, and in particular relates to a new method for video retrieval through video clips. Background technique [0002] Intelligent information processing technology with image and video information processing as the core provides a powerful tool for people to make full use of multimedia information. However, with the emergence of large-scale image and video databases, the amount of data that information systems have to deal with has increased significantly. How to effectively index and retrieve these databases has become one of the bottlenecks in the field of multimedia information processing. Content-based Image and video retrieval methods have opened up new ways for people to solve this problem. Video segment retrieval is the main method of content-based video retrieval. It refers to finding all similar segments from the video library given a query segment. [0003] Existing fragment retrie...

AI Technical Summary

Problems solved by technology

This method can perform precise retrieval and similarity retrieval at the same time, but its problems are: (1) It only considers the number of similar shots of two clips, but does not consider the impact of many-to-many shot similarity (granularity) on the overall similarity , therefore, even if all shots of segment Y are similar to only one shot of segment X, Y will be considered similar to X; (2) The proposed assumption does not hold, that is, the similar frames inside the two shots do not necessarily correspond in time order Similarity; (3) The similarity of the shots is judged based on the longest sequence of similar frames of the two shots. This kind of comparison based on each frame makes the retrieval speed of the clips relatively slow.

Images