Infrared Image Super-resolution Algorithm Research

For decades, Infrared image and super-resolution reconstruction (SRR) have been hot research topics in the field of digital image processing. The paper mainly makes a study of infrared image’s super-resolution reconstruction. Super-resolution reconstruction is a resolution enhancement technology for estimating a higher resolution image from a sequence of observed lower-resolution, which make full use of the similar, but not the same information among each observation.The thesis carries on the research about the image formation model of SRR, which is a typically ill-posed and inverse problem. So it needs to put forward an effective regularization term to achieve a unique and stable solution. This thesis focuses on the super-resolution of image sequence and it is carried out in the following aspects:Infrared image preprocessing, image registration and the regularized reconstruction algorithm. Since the accuracy of the registration algorithm results play a key role in SRR, the paper proposes an algorithm that combines the LK optical flow method and Gaussian Pyramid method and achieve a motion estimation of sub-pixel accuracy. And the thesis puts forward a new model of adaptive regularization method, researching into the choices of the regularization parameter and the contribution of each image. All those improvements can reduce computational-complexity and avoid error.Experimental results of the revised method illustrate the advantages not only for qualitative calculation but also for quantitative analysis. This experimental images show the new method is effective and better than others. And the new algorithm can highlight the details of the LR images and improve the infrared image’s resolution.