Incremental principal component pursuit for video background modeling

摘要

An incremental Principal Component Pursuit (PCP) algorithm for video background modeling that is able to process one frame at a time while adapting to changes in background, with a computational complexity that allows for real-time processing, having a low memory footprint and is robust to translational and rotational jitter.

主权项

1. A computer system method for video background modeling including an incremental principal component pursuit algorithm comprising a low computational cost executed by a processor, the method comprising the steps of:
providing the incremental principal component pursuit algorithm with the low computational cost O(14·m·r)+O(15·Nc·Nr(2·log2(Nr)+log2(Nc)) per frame per inner loop; receiving an input video sequence comprising a plurality of frames; estimating a low-rank approximation of a background of the input video sequence; segmenting foreground objects from background objects appearing in the input video sequence by processing the input video sequence in an incremental fashion to obtain a processed input video sequence, wherein a computational complexity of each frame of the input video sequence is O(8·m·r)+O(r3), where m=Nr (frame rows)×Nc (frame columns)×Nd (frame depth or channels), and r is the rank of a matrix that represents the background; and displaying the processed input video sequence to obtain an output video sequence.