Figure 10
From: Video analysis-based vehicle detection and tracking using an MCMC sampling framework
Illustration of sampling process for different example images. From left to right, images correspond to the (1) original image, (2) rectified domain, (3) appearance-based vehicle probability map, B m , (4) motion-based vehicle probability map, and (5) tracking results. The sampling process is illustrated in images (3) and (4): accepted and rejected particles are painted in green and red, respectively. Images (2)-(4) are zoomed for better visualization of the sampling process. Images in (a) illustrate a normal sampling scenario, while images in (b) and (c) show how combined sampling is able to overcome bad (b) motion-based and (c) appearance-based measurements.