| Video has become one of the most important media access to information in modern society. In order to satisfy the public’s need for high quality video, currently manufacturers have introduced high-definition even ultra hd video playback system, However, due to lack of high resolution data sources, It is urgent need to study high resolution video super-resolution reconstruction method based on low resolution video. Spatial resolution is one of the most important index of image quality, and the image super-resolution technology using software method without changing hardware conditions provides a feasible way for the existing low resolution video conversion to high resolution video. This paper studies the reconstruction from standard video to high-definition video, which has a great significance in theory and practical value.This paper proposed a super-resolution algorithm based on the wavelet feature sparse representation for standard video images. Firstly,At the training stage of the samples,in view of the high and low resolution sample ambiguity problem, we use the samples selection through calculating the gray-level uniformity and edge density to eliminate ambiguity. Besides, Combined image wavelet decomposition and sparse representation, this algorithm could rebuild the high frequency detail targeted. Image wavelet decomposition is used to divide the image of the high frequency into horizontal details, vertical details and diagonal direction details, we obtain three directions of details dictionaries with the dictionary learning algorithms. Finally, in the reconstruction phase we join in the process of super-resolution reconstruction pretreatment using Lucy-Rechardson image restoration algorithm and median filtering to improve the quality of the source sd video images, meanwhile,we use non-local means denoising algorithm as post-treatment to get a better visual effect. The actual standard definition video experiment results show that the proposed method has good performance compared with the current relatively popular super-resolution algorithm based on sparse representation. |
PDF Full Text Request