Multiple-frames super-resolution for closed circuit television forensics

CCTV forensics is a very challenging task as video files seized from a CCTV DVR can be in a detrimental state with bad lighting ambience, noise, blurring due to optical settings and motion, compression artifacts, interlaces and low-resolved frame size to name a few. This always results in the difficulties in analyzing subjects in the video which are subjected to investigation. We present a multiple-frames Super-Resolution technique by combining a sequence of video frames of a subject in order to create a super-resolved frame of the subject with increased resolution and clarities. For that purpose, we used Projection onto Convex Sets (POCS) method for the super-resolution. For estimating the shift and rotational parameters of frames for the POCS we used Keren for that purpose. This process is tested with real footage of CCTV video which we did simulation on a crime scene. From the experiments we conducted, the results found are far exceeding in quality than conventional image resampling methods.