Study On Video Images Super-Resolution Restoration Technique

Video image super-resolution (SR) restoration technique can generate higher resolution restored images from low resolution video images, remove the noise and the blurs caused by sensors and optics, and make the restored image break through its original resolution limitation. In the thesis, the investigation and experiments are carried out on the video image super-resolution restoration technique. Most work is focused on the motion estimation, the improved Projection onto Convex Sets (POCS) SR reconstruction algorithm considering motion blurs, and the Approximate PSF Examining (APEX) algorithm. The main work and innovations in the thesis are listed as follows.1. Discussion is given on the development history and classification categories of the video SR restoration technique. Analysis is made on the disadvantages and shortcomings of current SR reconstruction algorithms, and explanation is illustrated on the significance of the thesis.2. Analysis and comparison is made on the two common motion estimation algorithms, optical flow and block matching, as well as their application situations and limitations. Experimental results show, the optical flow algorithm is suitable for small local motions, while the block motion algorithm can acquire accurate motion parameter in the case of large local motion or global motion.3. Based on the several common block motion estimation search strategies, a linear parallel search strategy is proposed to reduce the computation complexity, which integrates the linear search and the parallel search. Experimental results demonstrate that the new strategy can efficiently improve the matching speed and ensure the precision of motion estimation simultaneously.4. Deep investigation is given on the whole video imaging model and the POCS algorithm, which consider motion blurs caused by the nonzero aperture time. The computing expression for the PSF in the video imaging model is deduced, and an improved POCS algorithm is proposed to reduce the oscillation effect. Experimental results prove that the improved POCS algorithm considering the motion blurs can effectually reduce the oscillation effect and advance the image spatial resolution.5. Based upon the deep investigation on the APEX algorithm, a new PSF computing method is proposed, which utilizes frequency domain limitation and fitness errors balance to acquire more accurate PSF model. Experimental results show that the improved APEX algorithm can obtain optimal restoration results with enhanced stability and applicability.