Key frame extraction method based on analysis sparse representation

Abstract

The invention discloses a key frame extraction method based on analysis sparse representation, comprising the following steps: converting a video into matrix representation, each column of a video matrix being each frame signal of the video; the method comprises the following steps of: designing an analytical sparse coding model based on minimax connectivity openalty (MCP) sparse regularization, and designing an analytical sparse coding model based on minimax connectivity openalty (MCP) sparse regularization; taking a video matrix as an original signal, taking a transposed matrix of the video matrix as an analysis dictionary, and substituting the analysis dictionary into the analysis sparse coding model; calculating a sparse coefficient matrix by analyzing a sparse coding algorithm, wherein a non-zero line of the sparse coefficient matrix represents an index of the key frame; and selecting the corresponding video key frame according to the index of the key frame. According to the method, the compression ratio of the selected key frame is improved, the calculation complexity is reduced, and the extraction speed of the key frame is improved. In addition, through verification of many challenging real world scenes, the method has higher extraction efficiency compared with a traditional key frame extraction method.

Description

technical field [0001] The invention relates to computer vision key frame extraction technology, in particular to a key frame extraction method based on parsing sparse representation. Background technique [0002] Artificial intelligence is increasingly filling our lives, and research on humanoid robots can assist humans in life or replace humans in harsh environments. As an important part of humanoid robots, computer vision gives robots the ability to perceive surrounding information. The development of computer vision has further expanded the breadth of robot perception accuracy, making it easier and more accurate to handle the next task. Video processing is an important topic in computer vision. Due to the huge amount of video data and a lot of noise, how to effectively extract the key frame information of the video is very important. The video signal is composed of connected frames, and the time redundancy is particularly large, that is, the inter-frame correlation is...

AI Technical Summary

Problems solved by technology

However, the existing sparse constraints cannot extract effective key frames, and are basically based on traditional synthesis models, which do not consider the structural information of video frames in the sparse coefficient matrix. Therefore, it is necessary to develop more effective algorithms to obtain better video key frames. frame is necessary

Images