METHODS OF IMAGE RESTORATION FOR INCOHERENT AND COHERENT SYSTEMS (DECONVOLUTION, SIGNAL RECOVERY, DEBLURRING, ENHANCEMENT)
Abstract
Satellite-based multispectral imaging systems have been in operation since 1972. The latest in the Landsat series of sensors was launched in March 1984. One of the system parameters of interest is resolution and in this thesis results in estimating the actual overall resolution after launch are discussed. Scene structures such as field edges are used with numerical estimation procedures to predict resolution. A discriminant function, based on statistical methods, is developed for the selection or rejection of the scene elements to be used. Different approaches are considered for processing the data used in the estimate of the LSF. For estimating the 2-dimensional PSF a parametric model using a finite number of edges in different orientations is presented. Knowing the PSF a restoration method is developed which, if desired, could be easily incorporated into the existing software used for processing the Landsat imagery. The restoration algorithm is a generalization of the method of compensating the Composite Point Spread Function (CPSF) using FIR filters, with the addition of direct constraints over the side-lobe levels present in the CPSF. Depending on the PSF and sampling rate, excellent results were obtained in controlling the side-lobe levels. In some cases they are reduced below the quantization level of the system, making them unnoticeable. An extension of the restoration method to coherent image formation systems is discussed. Results in enhancing TM Landsat data are presented.
Degree
Ph.D.
Subject Area
Electrical engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.