Local Structural Similarity Based Video Super Resolution Algorithm

Image and video are the very important research in the area of computer science, and is very hot in recent years.With the developing of the technology of image and video processing, there is much higher level for the quality of image and video.High resolution image and video contains more details and information,which provides more detail for the processing of image and video.There are two methods to get the high resolution image and video.The one is to improve the quality of hardware which can capture video,and the other one is to do super resolution.But it is impossible to change the sensor of capturing video device smaller all the time, and replacing the sensor costs so much.So it is the only choice to get high resolution image and video from the super resolution on image and video.Video super resolution aims to reconstruct the video sequence with the high spatial and temporal resolution by integrating the information of low resolution video sequence from the same scene.Video super resolution contains temporal resolution and spatial resolution.The temporal resolution needs to restore the lost temporal detail and information, and the spatial resolution is considered as the image and video super resolution restoration.This paper discribes the developing process of video super resolution and technologies existed, and then points out the advantages and disadvantages of the existed super resolution technologies.This thesis proposes a new video super resolution algorithm based on local structural similarity.Any frame is selected to be proceeded and some other frames are choosed to be searched in some manner.The object moves forward and backward in the video because of the moving of camera.So the resolution of the same object is different between different frames.We can reconstruct the smaller object information using more details and information of the larger object.At the same time we generate the image sequence of low resolution through local structural similarity existing in any single frame and among frames.Through the IBP algorithm we can restore a high resolution image,and then output it.We find that the proposed algorithm can keep the edge sharper and hold the detail through analyzing the test result and comparing with the image getting from the other interpolation algorithm.This thesis describes the important technologies and the test system which runs the algorithm.