Study Of Super-Resolution Image Enhancement Technique In Surveillance Video

It is not conducive to distinguish the detail feature of objects, due to the low quality of collected surveillance video images resulting from the environmental changes, down sampling and other causes. However, it has not only high cost but also long cycle by hardware method to improve the image quality. So there is large important value of research and application by the software algorithm of super-resolution to reconstruct video image.This paper reviews four classic interpolation algorithms, and proposes an improved diamond-edge interpolation algorithm combined with bilinear interpolation and edge-orientation covariance. After a depth analysis for motion estimation, it establishes the optical the constraint equation of flow vector. Thus the obtained model has variability by using an adjustable weight coefficient instead of weight factor. Then a method to extract qualified reference frame as the static frame was proposed for the selection of reference frame for improving the precision of motion estimation. Finally, it proposes an iterative regularization algorithm of image sequence based on bilateral filtering operator to remove the noise while preserving edges. The bilateral filter is a constraint to improve the quality of reconstructed image by using the steepest descent method for solving the corresponding energy functional. The bilateral filtering operator term can not only suppress the amplification of the noise effectively but also maintain the important detail information of an image’s edge. The experimental results show that the reconstruction algorithm has very strong robustness on the simulated data and the real environment.