Research On Image Super-resolution Reconstruction

Image super-resolution reconstruction is the image processing technology to produce the image of the high-quality, high-resolution from the low quality, low-resolution in the single or multi-images. In the thesis, it is elaborated and the algorithms involved are focused on. Briefly speaking, the main work includes one image reconstruction from one image and several images (or image sequence), image sequence reconstruction from image sequence, the criterion for super-resolution, etc.The research is listed as follows.First, the research background and the situation at home and abroad of the art are reviewed, with its presence in various fields of military and civilian application indicated. Furthermore, an overview of current methods employed to the SR reconstruction is given, the fundamental skill, difficult issue and future development are all discussed. Based on these, the relation and the difference between the SR reconstruction and the technologies related are further such as image restoration, resolution, image enhancement and blur, etc. clarified.For the problem of one image super resolution reconstruction from one image, a novel method is developed. A flaking method is suggested in the design of the algorithm; a more efficient approach is presented in the implement of the algorithm, the method avoids the operations of huge matrix. Compared with the traditional method, it is efficient and low complexity.As for the problem of one image super resolution reconstruction from one image, we develop a novel method. A flaking method is suggested in the design of the algorithm; a more efficient approach is presented in the implement of the algorithm, our method runaround the operations of huge matrix. So, comparing our method with traditional method, the former is efficient and low complexity.As far as super-resolution construction from image sequence to single image is concerned, a new B-spline interpolation method based on the image sequence is proposed by using the additional data in the sequence of existing mobile-related sub-pixel to improve the resolution. Besides, a integrated MAP/POCS hybrid approach is also proposed based on probability theory and the collection methods, with the derivation of the equations of the algorithm presented and spin-off of a single image reconstruction algorithm method applied to the specific algorithm implementation.Some key problems including the model of image degradation, the estimation of blurring function and the method for big size equation system in super resolution reconstruction are studied deeply. For the high and low SNR, the frequency method and error-parameter estimative method are discussed; for the big size system, an effect pre-process method is considered.In the last part of the thesis, some methods for the criterion of super-resolution are presented and the method of profile density curve is introduced. We also proposed some ideas as open problems for this problem.