Techniques of deconvolution, interpolation and super-resolution for high-resolution image reconstruction
Li Chen · 2006
This thesis investigates how to produce a high quality, high-resolution image from low quality, low-resolution images.In many visual applications, high quality images are desired but may fail to be obtained because of some degradation factors.Several images, which suffer from the degradations but consist of the overlapping content of a scene, are used to produce a single image of superior quality.Using high-resolution reconstruction, it is possible to restore high-frequency content, reduce noise, and even increase spatial resolution when hardware modification is unrealizable.The generic name of high-resolution image reconstruction covers related subjects of deconvolution, interpolation, and super-resolution.Image deconvolution mainly deals with deblurring from blurred and noisy images, while the major goal of interpolation and super-resolution is to increase spatial resolution from the aliased images.In the first part, a series of algorithms are proposed to solve different problems encountered in blind image deconvolution.In Chapter 2, several efficient discrete spatial techniques for blur support identification are derived and analyzed.A soft modeling algorithm is proposed to generate the manifold parametric blur models and determine the final blur estimate in Chapter 3. We attempt to address blind deconvolution by assessing the relevance of parametric blur information, and incorporating the knowledge into the parametric double regularization scheme.Further, an iterative algorithm based on multichannel recursive filtering is proposed to address multichannel image deconvolution.In the second part, the image interpolation is formulated as a regularized least squares ATTENTION: The Singapore