Wavelet-based image and video super-resolution reconstruction.

Super-resolution reconstruction process offers the solution to overcome the high-cost and inherent resolution limitations of current imaging systems. The wavelet transform is a powerful tool for super-resolution reconstruction. This research provides a detailed study of the wavelet-based super-resolution reconstruction process, and wavelet-based resolution enhancement process (with which it is closely associated). It was addressed to handle an explicit need for a robust wavelet-based method that guarantees efficient utilisation of the SR reconstruction problem in the wavelet-domain, which will lead to a consistent solution of this problem and improved performance. This research proposes a novel performance assessment approach to improve the performance of the existing wavelet-based image resolution enhancement techniques. The novel approach is based on identifying the factors that effectively influence on the performance of these techniques, and designing a novel optimal factor analysis (OFA) algorithm. A new wavelet-based image resolution enhancement method, based on discrete wavelet transform and new-edge directed interpolation (DWT-NEDI), and an adaptive thresholding process, has been developed. The DWT-NEDI algorithm aims to correct the geometric errors and remove the noise for degraded satellite images. A robust wavelet-based video super-resolution technique, based on global motion is developed by combining the DWT-NEDI method, with super-resolution reconstruction methods, in order to increase the spatial-resolution and remove the noise and aliasing artefacts. A new video super-resolution framework is designed using an adaptive local motion decomposition and wavelet transform reconstruction (ALMD-WTR). This is to address the challenge of the super-resolution problem for the real-world video sequences containing complex local motions. The results show that OFA approach improves the performance of the selected wavelet-based methods. The DWT-NEDI algorithm outperforms the state-of-the art wavelet-based algorithms. The global motion-based algorithm has the best performance over the super-resolution techniques, namely Keren and structure-adaptive normalised convolution methods. ALMD-WTR framework surpass the state-of-the-art wavelet-based algorithm, namely local motion-based video super-resolution.