A Research On Multiframe Images Resolution Enhancement

Superresolution refers to image processing algorithms which produce high quality, high resolution (HR) images from a set of low quality, low resolution (LR) images. In this dissertation, some main issues of super-resolution including the preprocessing, the superresolution algorithms, the postprocessing and the application of superresolution are researched and some significant results are achieved. The major work implemented are presented as follows :1. A comprehensive review of the basic theories and techniques of superresolution is addressed . The process of acquiring images and the major factors affecting images' quality are analyzed briefly and the characteristics of some commonly used geometric transformations models and their inherent relationship are discussed.2. Based on the feature of smooth geometric distortion and high accuracy feature point extraction and matching in digital photogrammetry , a self-adaptive high accuracy image registration algorithm is proposed, and the region filling algorithm is implemented to improve the efficiency of image registration, which lays a good foundation for remote sensing images superresolution.3. Based on the multiresolution analysis with orthonormal wavelets, the wavelet superresolution algorithm for the more general camera motion model is analyzed. The implementation of this algorithm to the superresolution for remote sensing image is a great progress.4. The widely-used Delaunay triangulation in computer graphics is introduced into image superresolution and a MDTHR algorithm is proposed. The experiments show it an algorithm of good performance and which is easy for data updating..5. Superresolution for interlaced remote sensing images is researched and the principles of imaging, degradation model and reconstruction process are analyzed. Furthermore, some key techniques for SPOT5 supermode images processing are researched.6. The combined ridge-stein estimator (CRS) is presented to improve the accuracy and robust for linear pushbroom imagery exterior orientation Based on the positioning theory in photogrammetry, the positioning accuracy of superresolution reconstructed image is quantitatively evaluated, the experiments show that its positioning accuracy improves obviously.7. The Richardson-lucy(R-L) algorithm for image restoration is researched and the acceleration and noise dampening of R-L algorithm are discussed. This algorithm is applied for the postprocessing for images reconstructed from spatial interpolation and satisfactory results is obtained.8. The basic theories and procedure of the MAP method and POCS method in spatial domain are summarized. These methods are applied for the superresolution for remote sensing images successfully and better superresolution reconstruction results are obtained. It is a progress made in applications for remote sensing image superresolution.9. The theories and procedures of superresolution via alias removal are analyzed and some typical experiments are performed, the correctness and feasibility of this method are proved in theory and practice.