Video foreground and background separation method based on robust low-rank sparse decomposition

Abstract

The invention discloses a video foreground and background separation method based on robust low-rank sparse decomposition, and the method comprises the steps: converting a to-be-processed video into atwo-dimensional matrix M with the size of m rows and n columns, m being the length-width product of each frame of the video, and n being the number of frames of the video; inputting the two-dimensional matrix M into a pre-constructed model of a video foreground and background separation method based on robust low-rank sparse decomposition, and outputting a low-rank matrix B corresponding to a video background, a sparse matrix F corresponding to a video foreground and a noise matrix G. According to the robust low-rank sparse decomposition-based video foreground and background separation method, a generalized non-convex kernel norm is adopted as a rank function of a model; a structured sparse induction norm is adopted as a 10 norm of the model; and a noise item is added. The method has theadvantages that compared with a traditional video foreground and background separation method, the robustness is good; the noise resistance is high; and the foreground and background separation effecton a noisy video is good.

Description

technical field [0001] The invention relates to a video foreground and background separation method based on robust low-rank sparse decomposition, and belongs to the technical field of video processing. Background technique [0002] Video foreground and background separation technology is a technology to extract key information from massive video data. It is the core step in the video processing process and plays an important role in traffic control, social security and signal processing. In the past ten years, people have carried out extensive research around this technology and achieved remarkable results. Initially, researchers used Principal Component Analysis (PCA) for video background modeling, which has been widely used in image recognition, image feature compression, face recognition and video foreground and background separation. However, this method is robust The performance is not high, the anti-noise ability is relatively weak, and it is easy to lose some key in...

AI Technical Summary

Problems solved by technology

Initially, researchers used Principal Component Analysis (PCA) for video background modeling, which has been widely used in image recognition, image feature compression, face recognition and video foreground and background separation. However, this method is robust The performance is not high, the anti-noise ability is relatively weak, and it is easy to lose some key information in the video, which will lead to poor foreground and background separation of the video; the low-rank sparse method is to decompose the video into sparse components and low-rank components, and convert them into optimized method, which has greatly promoted the development of video foreground and background separation methods, especially the principal component analysis algorithm (Principal Component Pursuit, PCP) based on low-rank sparse decomposition. Although this method has greatly improved the effect of video foreground and background separation, but The effect of video processing with serious noise corrosion is not good

Images