Super-resolution Image Reconstruction Algorithm Study Based On The Multi-scale Geometric Analysis

Multi-scale geometric analysis is the research hot spot in the field of scientific research, the digital image processing, the video sequence processing and many other core technologies. As the inherent limitations of hardware devices in the process of image, we are unable to obtain high resolution images. By improving the hardware equipment, we not only need to spend a lot of money to improve the resolution of the image, but it is difficult to solve some technical problem in the short term. So the method of software has become the research hot spot. Super-resolution reconstruction technique had been widely used in surveillance, military, bio-medical, remote sensing and other fields. The thesis mainly works and achievements are as shown below:Chapter one: this chapter introduces the history of the technology of super-resolution reconstruction, and some applications in the real life. In addition, the aim and intention of the paper is introduced.Chapter two: this paper mainly discusses the mathematical model of image reconstruction and the algorithm of image super-resolution reconstruction.Chapter three: Some basic theories and principle of wavelet transform and contourlet transform are introduced. Comparing for two kinds of the denoising method, we choose the contourlet denoising. The numerical experiments also show that the method is better than the wavelet denoising.Chapter four: this chapter studies the single frame super-resolution image reconstruction.Meanwhile, a method of super-resolution image reconstruction based on wavelet transform and edge-directed interpolation is put forward. Finally the numerical experiments also show that the method is better than the traditional interpolation method.Chapter five: Firstly, for multiple low resolution images registration, We use the Keren method. Then, on those high resolution grids, we have the interpolation to obtain one initial estimate. Last, we apply wavelet transform and edge-directed interpolation method to improve the resolution of the image. As the meanwhile, we use the contourlet denoising method to reduce noise interference. Finally the numerical experiments also show that the method is better than the traditional interpolation method.Chapter six: A brief summary of the present work is given. Moreover, the further research direction is pointed out.