A Fast And Robust Single Image Super Resolution Reconstruction Method

Image super resolution reconstruction is to reconstruct a high-resolution image from a single or multiple low-resolution images of the same scene.In many cases,we cannot(afford to)capture multiple images in the same scene,such that single-image super-resolution(SISR)has a wider range of application requirements.However,SISR reconstruction is more ill-posed and more challenging than multiple-image super-resolution reconstruction because the latter can make use of more information than the former.Due to the complexity of the model and the corresponding algorithm,the existing SISR methods are often very time-consuming and cannot cope with complex real-world super-resolution reconstruction requirements.Therefore,the study of SISR reconstruction has important theoretical significance and application value.In this thesis,firstly,the concepts,models and algorithms of image super-resolution reconstruction,and the image quality assessment index are introduced.The methods of image deblurring and image denoising are expounded and compared.The feasibility of transforming the image super-resolution reconstruction problem into blind image deblurring(BID)is analyzed.Then a super-resolution algorithm model based on BID is proposed,and the algorithm is implemented in the form of alternating direction method of multipliers.The proposed algorithm based on BID is always faster than existing SISR reconstruction algorithms since the BID is usually easier to be solved than the SISR.Finally,an improved super-resolution image model is proposed by combining the adaptive median filtering algorithm,such that the proposed single-image super-resolution algorithm is robust to noise.The effectiveness of the proposed algorithm are verified by comparing with the classical SISR algorithms.