Use this URL to cite or link to this record in EThOS: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.576422
Title: Super-resolution image reconstruction from low-resolution images
Author: Nasir, Haidawati Mohamad
Awarding Body: University of Strathclyde
Current Institution: University of Strathclyde
Date of Award: 2012
Availability of Full Text:
Access through EThOS:
Access through Institution:
Abstract:
The thesis addresses the problem of obtaining high-resolution image from a set of one or more low-resolution images. The thesis focused on three building blocks of super-resolution algorithms i.e., image registration for super-resolution, image fusion for super-resolution and super-resolution image reconstruction. These three parts are addressed separately and singular value decomposition-based fusion is introduced before performing interpolation or single-image super-resolution. An accurate image registration is crucial for super-resolution. An image registration approach for super-resolution based on a combination of Scale Invariant Feature Transform (SIFT), Belief Propagation (BP) and Random Sampling Consensus (RANSAC) is described to automatically register the low-resolution images. The results have shown effective for the removal of the mismatched features in the image. A novel SVD-based image fusion for super-resolution is developed for integrating the significant features from low-resolution images. The SVD-based image fusion is shown to enhance the super-resolution results. The implementation of a novel interpolation method based on a linear combination of the bicubic interpolation and their first-order derivates and the use of first-order difference equation to extract the features from the low-resolution images are described and shown to improve the method of single image super-resolution using sparse representation. The proposed method has shown to reduces the computational time and enhance the prior estimation of the high-resolution image as well as the final super-resolution results. The performance of the algorithms is evaluated using synthetic sequences and also on real sequences subjectively and objectively.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.576422  DOI: Not available
Share: