Based On Wavelet And Super-resolution Reconstruction

Due to the pixel size limitation of the remote sensors and the interference of the atmosphere on sensors, it is very difficult to acquire a remote sensing image of high resolution but less distortion (i.e., noised images). The super resolution reconstruction can overcome the inherent resolution constraint of current imaging systems, and this feature leads to a significant improvement of the related technologies in many applications such as video monitoring, medical, remote sensing and object tracking.In this thesis, I focus on the investigation of the super resolution reconstruction using image wavelet transform. An existing image of high-resolution is decomposed into wavelet coefficients which are used to describe image edges in distinct directions, and these coefficients are then applied to other images of low resolution to reconstruct that of high-resolution. As the literature review of this thesis, I present a brief introduction about the remote sensing imagery and some key issues with the highlight on imaging resolution. And then I systematically explored the wavelet transform and its multi-resolution features, in addition, a detail analysis of the Mallet algorithm is carried out for obtaining the wavelet coefficients. Subject to the rule of the minimum absolute error, I propose an approach to estimate the Wavelet coefficients. With this edge information of low solution images at high-scale decomposition, the high resolution images are then generated. Furthermore, a noise filtering, which is based on the Wavelet soft thresholding, is also taken into account in order to tackle the problem in terms of too much noise in remote sensing images.