Figure 9
From: Object detection oriented video reconstruction using compressed sensing
Object detection results of pedestrian video sequence. (a) Original video sequence. (b) Reconstructed foreground using the proposed algorithm. (c) Reconstructed foreground using RPCA. (d) Reconstructed foreground using GMM. (e) Reconstructed foreground using modified RPCA (the manual postprocess result of RPCA). The sampling rate of our algorithm is 40% and four frames, i.e., 1st, 8th, 12th, and 15th frame are randomly selected.