Jitter invariant incremental principal component pursuit for video background modeling on the TK1

While Principal Component Pursuit (PCP) is currently considered to be the state of the art method for video background modeling, it suffers from a number of limitations, including high computational cost, batch operating mode, and sensitivity to camera jitter. In this work we present a real-time, CUDA-aware C / CUDA C implementation of a novel and fully incremental PCP algorithm for video background modeling that can also deal with rigid transformation jitter. Our implementation has a computational complexity that allows (TK1 platform) a processing frame rate throughput (jittered video) of 2.6 and 0.4 f.p.s. for color videos of 640 × 480 and 1920 × 1088 respectively.