Research On Single Video Super-Resolution Reconstruction By Using Spatial-Temporal Correlation

Single video super-resolution technique has been widely used in the field of video processing,such as surveillance video processing,biomedical video processing,video enhancement and restoration,computer vision,virtual reality,augmented reality and other emerging areas.However,the existing single video super-resolution algorithms still have some shortcomings.The existing single video super-resolution algorithms do not fully exploit the correlations in intra-frames and inter-frames,and the modeling of the low-resolution video sequence is incomplete.Therefore,it is still possible to improve the results obtained by the above methods.In addition,most algorithms reconstruct the non-key frame based on key frames,and this makes the quality of the result strongly rely on the accurate motion estimation(ME).Therefore,inaccurate ME leads to the annoying artifacts.To solve these problems,in this thesis,we study the suitable single video super-resolution reconstruction algorithms,and the following results are achieved:(1)In order to fully exploit inter-frame and intra-frame correlation in video,a single video super-resolution algorithm based on reconstruction is proposed,and an optimal solution model is designed.When exploiting the spatial correlations,the non-local means model is used to get the non-local structural property and the total variation model is utilized to get the local structural property.In order to exploit inter-frame correlation,optical flow method is applied to perform inter-frame estimation.Finally,to solve the established optimization problem,a split-Bregman method-based iteration is proposed.The experimental results demonstrate that the effectiveness of the proposed algorithm.Compared with other existing algorithms,the proposed algorithm is able to achieve better subjective and objective results in different sequences.(2)Most of the existing single video super-resolution algorithms are based on accurate motion estimation.When processing video sequences,if the motion is estimated inaccurately,the super-resolution results will be distorted.In order to solve the problem that the existing single video super-resolution algorithm depends on the accurate motion estimation,which leads to the problem that the super-resolution results is in thrall to the accuracy of motion estimation and the characteristics of the processed video,a three-dimensional non-local mean constraint model is proposed and used for single video super-resolution reconstruction.The three-dimensional non-local means is used to explore the inter-frame and intra-frame non-local structural property,and the total variation model is utilized to describe the local structural property.Finally,to solve the established optimization problem,a split-Bregman method-based iteration is proposed.The experimental results demonstrate the effectiveness of the proposed algorithm.Compared with other algorithms,the proposed one is able to achieve better subjective and objective results.At the same time,the proposed algorithm does not rely on accurate motion estimation,and has a wide range of applications.