Research On Video Sequence Based Super-resolution Image Reconstruction Technique

Super-resolution image reconstruction technique based on video sequence is a technique that constructs a high-resolution image using redundant information between each other under the premise of low-resolution image sequences which have sub-pixel shift to show detail information better. At present, super-resolution image reconstruction technique has been widely applied in the field of image reconstruction, such as military remote sensing, medical imaging, video surveillance, image compression, high-definition television and so on. Therefore, it has become one of the most active research topics in the field of image reconstruction.Firstly, this paper has summarized the mathematical model and reconstruction algorithms on image reconstruction, analyzed and compared the advantages and disadvantages between the existing super-resolution image reconstruction algorithms. Moreover, this paper has introduced the key techniques in details.Secondly, this paper has focused on Horn-Schunck algorithm based on optical flow registration and made improvements. The paper has used a variable weight instead of the constant, thus the calculating model has variability. The comparison of experimental results and data between the improved algorithm and the original has been done under MATLAB 2009a platform.Thirdly, this paper has studied the technology of image interpolation. It has introduced a robust determination and an adaptive structure function instead of the fixed determination and function based on normalized convolution. The experimentations have been proved that the robust determination can reduce image noise effectively and the adaptive structure function can sharp the reconstructed image and void jitter and jaggies.Finally, the normalized convolution algorithm of robust adaptive structure after improved Horn-Schunck optical flow registration has been experimented under MATLAB 2009a platform. The experimental results of the noise and noise-free image reconstruction have shown the superiority of the algorithm in data comparisons of visual quality, root mean square error and peak signal to noise ratio. Moreover, in combination with the practical project, the result of real image reconstruction also has proved that our algorithm can improve the image resolution without changing hardware conditions, so it can achieve the effect of super-resolution image reconstruction.