Research Of Decoding-Video Resolution Enlargement Algorithm Based On Reference Frames

With the fast development of digital technology, network technology and communication technology,the application of digital video technology is more and more extensive.It covers many fields of the people's living and work.While playing the video with a very big screen and a very high display resolution, even if the quantity of the video is very good, but when the people play the video with a big window, if the resolution of the video is not enough big, the block and serration distortion will be more evident,and the distortion also makes the people's eyes unconformable.To solve this problem, we should raise the resolution of the video.This thesis analyzes and studies the technology of the decoding-video resolution enlargement based on researching the method of the domestic and international existing video resolution enlargement.The thesis puts forward a method of decoding-video resolution enlargement based on reference frames in allusion to the shortage of the traditional method of video resolution enlargement and by combining the characteristics of the video compression coding and decoding system.To enlarge the decoding I frame,firstly,we abstract the figure of the edge and detail for I frame,and then we reconstruct again the decoded and reconstructed I frame enlarged by interpolation according to the characteristics of its figure of the edge and detail.Thereby, we attain our purpose to enlarge the resolution of I frame.Aiming at the high computing quantity of the enlargement algorithms in the picture operating field, our interpolation enlargement algorithm adopts a simple algorithm.To enlarge the decoding prediction-frame (P frame and B frame),we utilize the temporal and spatial correlations among the frames of continuous video,and their edge-detected results are based on the results of the I frame.When enlarging the prediction-frame,to the inter prediction model,firstly we take block as the unit and enlarge the residual block by bilinear interpolation enlargement algorithm.Then,we reconstruct and enlarge every block by adding every enlarged residual block to its reference block.And to the intra prediction model, we also take block as the unit and enlarge the prediction-frame by using an"enlarged intra prediction model". So we can reconstruct and enlarge every prediction-frame picture.The experiment results show that our algorithm has evident effect on improving the subjective quality of both every frame image and the entire video,and the block and serration distortion are also reduced largely when video is played enlargedly.