Regularization Super-resolution Reconstruction Algorithm

As people's requirements to image resolution improved, How to improve the spatial resolution of the image in the existing level of hardware becomes an urgent need to solve, image super-resolution reconstruction technique is proposed in this demand. Image super-resolution reconstruction technique reconstructs a high-resolution image from two or multiple low-resolution images which are degraded by warping,blurring and noising. The technique is applied to public safety field,digital TV domains, military in remote sensing, medical fielding,and so on, which has become one of the hottest topic in current image processing area.Generally speaking.image super-resolution reconstruction technical includes image registration.interpolation and de-blurring. The three steps can also be done jointly. The main work of this paper is summarized as follows:Firstly,the image super-resolution reconstruction of the research background,research dynamic and development direction at home and abroad are introduced. Then expound the meaning of the image super-resolution reconstruction and principled.The observation model of super-resolution, the evaluation method of image reconstruction and reconstruction steps are also introduced.Secondly, For the two important aspects of image super-resolution reconstruction process:image registration and reconstruction algorithms, we focus on the Kanade-Lucas-Tomasi (KLT) sub-pixel level image registration algorithm.To prove its superiority we compare with the conventional image registration algorithms. On this basis, regularization method in image super-resolution reconstruction is introduced. We do research on Tikhonov,Ll norm TV (total variation) and L1norm BTV(bilateral total variation) image reconstruction algorithms.Then we improve the LI norm BTV (bilateral total variation) image reconstruction algorithm based on the shift &add algorithm.Against, based on the Chinese characters characteristics,We improve the existing regular expressions so that it is more suitable for text image super-resolution reconstruction. Super-resolution reconstruction experiments prove the obvious effect of the improved algorithm.Finally, we summarize our work and give the directions of future research.