Super-resolution Remote Sensing Image Reconstruction Techniques

With the development of the Remote Sensing technology, a large amounts of images of the earth with multi-platform, multi-sensor, multi-phase, multi-spectral. In these data, there are considerable image data of the same area. But limitted by the manufacturing process of imagery equipment, these images are the undersampled of scene. Also, the motion between the earth and the platform, atmosphere disturbance and so on make the images blurred and degraded. It's the aim for many pepole to improve the image resolution. One of the proper approaches to the aim is the siginal processing based methods in software. It can recover the high resolution image from multiframe undersampled images.This dissertation first describes the feasibility and the ablity of restoration by analyzing the image processing. The complementary information between images becomes the main cause of restoration. The major work implemented in this dissertation is presented as follows:1. Give a brief introduction of the rgistration method based on the elementary segments of differetial rectification and the motion model estimate with conjugate gradient.2. Study the spectral de-aliasing method of image restoration in frequency domain, and offer the improvement with block processing and recursive rstoration to some problems. At the same time, the biased estimate theory is applied in the restoration processing. These improved steps increase the value in the use of data and decrease the time of computation.3. Discuss the wavelet interpolation method in spatial domain with the multiresolution anaysis, and give a model of multi-scale reconstruction from multiframes to superresolution.4. Develop a practical superresolution image restoration system , which includes some basic image processing and the superresolution of image.