A content-based video retrieval method and device

Abstract

The invention provides a content-based video retrieval method and apparatus, and belongs to the field of a computer searching technology. The content-based video retrieval method comprises the steps of performing retrieval on multiple hash tables formed by key frame codes according to any one target key frame code, and determining a key frame code matched with any one target key frame code; calculating a similarity score between any one target key frame code and each matched key frame code; calculating an overall similarity score between a target video and each matched video based on timing sequence information of the key frames according to the similarity score between each target key frame code and each matched key frame code; and taking a matched video corresponding to the overall similarity score which is greater than an adaptive score threshold value as a retrieval result. Video retrieval is performed according to the timing sequence information of the key frames; and different videos with similar partial contents can be distinguished according to the timing sequence relation, so that the accuracy in video retrieval is relatively high, and a relatively good effect in repeated video distinguishing is achieved.

Description

technical field [0001] The present invention relates to the technical field of computer search, and more specifically, to a content-based video retrieval method and device. Background technique [0002] With the rapid development of Internet technology and the continuous updating of video capture equipment and video editing software, the number of videos circulating on the Internet has exploded. A large number of videos are uploaded, downloaded and shared on various social applications every day. While people can easily obtain various video resources, they can also use various video editing software to edit videos, such as converting from simple video formats to video clips, multiple transformations and mixing, etc. As a result, many repeated or nearly repeated videos are stored on the social application server, thereby consuming a large amount of storage space. Therefore, how to retrieve the repeated or nearly repeated videos in the video database is the focus of people's...

AI Technical Summary

Problems solved by technology

But in fact, due to the different timings of similar scenes in different videos, the video corresponding to the key frame encoding of similar videos does not overlap with the target video globally. The above judgment results will cause the target video to be misjudged as duplicate or approximate duplicate video of

Therefore, the accuracy rate is lower when video retrieval

Images