Visual comparison on background subtraction results over three scenes of UCSD data set. From top to bottom: surfers, boats and birds. From left to right: (a) input frame, (b) saliency map generated by BMS, (c) ground truth, (d) proposed approach, (e) 3WD , and (f) RMAMR. The top 3 best algorithms (organized by rank) was chosen. The true positives pixels are in white, true negatives pixels in black, false positives pixels in red and false negatives pixels in green.

Visual results of SCM-RPCA over three scenes of MarDT data set. From left to right: (a) input frame, (b) saliency map with its temporal median subtracted (due to the high saliency from the buildings around the river), (c) low-rank component, (d) sparse component, (e) foreground mask, and (f) ground truth. The true positives pixels are in white, true negatives pixels in black, false positives pixels in red and false negatives pixels in green.