Algorithm Research Of Video Image Super-Resolution Reconstruction

Superresolution reconstruction of video image is used to obtain high resolution video image. It not only improves vision effect of video image, but also opens the door for analysis and process of video image, etc. So the research of superresolution recongstruction of video image has an important significance.First, this thesis introduced the background, significance and the current research situation of domestic and abroad of the Video image super resolution reconstruction technology. Then the video image quality decreasing process was modeled according to the camera imaging theory, video image imaging characteristics, ambient inference in imaging process and hardware problems. Current super resolution reconstruction algorithms were analyzed according to current research. As far as video image is concerned, super resolution reconstruction technology is high resolution video image through evaluating a series of low resolution video images. From the modeling process of video image, blur noise seriously affect the resolution of the video. Generally video image super resolution reconstruction of a low resolution image or video image must make the result not perfect. Therefore we should preprocess the original image or video image before super resolution reconstruction.This thesis used four low resolution images to reconstruct one high resolution image. The images came from the same scene with relative displacement. And the thesis also used three adjacent frames from the low resolution video image to reconstruct the middle frame. The innovation of superresolution reconstruction contains several parts: 1) improved the algorithm of deceasing additive noise according to the character of video image noise; 2) superresolution reconstruction is an inverse and ill conditioned problem, we constructed regularization parameters with priori knowledge to resolve the problem, and it made the reconstruction result better than before; 3) iterative algorithm was used to realize registration with high accuracy which improved the effect of video image superresolution reconstruction.This thesis use matlab to make program, and the video images which were used to research superresolution reconstruction came from camera or standard video database.