Study Of Video Image Multi-level Super-resolution Reconstruction Based On Sparse Representation And Dictionary Leaning

Video and image super-resolution reconstruction is an important research direction of video and image processing, as well as recent research hotspot. Image super-resolution reconstruction refers to the process of a single or multiple low resolution images reconstruct of a clear high resolution image, and has been widely used in recent years in video surveillance, medical image processing, HDTV and other fields.In the video image super-resolution reconstruction, the method based on sparse representation and dictionary learning reconstruction result is better than others. This method can recover more details information, both visual and numerical measure, can obtain good reconstruction results. But when it is used to reconstruct whole area of the video image, it takes a long time, rebuild efficiency is low.To solve the above problems, this paper presents a video image multi-level reconstruction method to improve the efficiency of video image reconstruction. In this paper, the author's major work and contributions are outlined as follows:1) Using hierarchical learning image super resolution reconstruction of video image in the interest area reconstruction. Hierarchical learning methods for image super resolution reconstruction when get the image edge and texture information is relatively abundant, the method used in video image is interested in the reconstruction of the area, can get interested in target better reconstruction result.2) This paper puts forward a kind of algorithm based on Snake to extract interesting region in video image. Using closed contour, Snake algorithm to detect moving targets will include precision closed contour minimum rectangular area as the interested region, can make be interested in area of moving targets in include regional minimum at the same time.3) Video image multi-level reconstruction method is proposed, interested area in video images of moving targets contained super-resolution reconstruction of hierarchical learning, for not interested area interpolation reconstruction, improves the existing super-resolution reconstruction method based on dictionary learning role all over the reconstruction of the image area brings problems for a longer time, lay a foundation for the real time video reconstruction.