Hardware and software developments for applied digital image processing
Thesis DisciplineElectrical Engineering
Degree GrantorUniversity of Canterbury
Degree NameDoctor of Philosophy
This thesis discusses some of the problems and practices of applying digital image processing to scientific and industrial problems. An interactive image processing system is described. This system has been found to be useful for the development of image processing commands and algorithms to solve image processing problems. An image capture subsystem based around a 576 x 385 element frame transfer charge coupled device is described. This system has a variable integration time, giving it considerable flexibility over more conventional cameras. The properties of the RANK filter are described. The usefulness of this nonlinear filter is demonstrated by its performance on a number of common image processing tasks. The edge detection properties of RANGE filters, which are the difference between the outputs of two RANK filters, are discussed. This family of filters has considerable flexibility in the type of output response obtained, but was found to be slightly more sensitive to noise than the commonly used SOBEL filter. A new iterative solution to the Fourier phase retrieval problem is described. This solution is less efficient than current solutions, but does prove the optimality of an operation common to existing iterative methods of solution. An approach to developing algorithms to solve image processing problems is described. The various commands that may be used at each stage are discussed. Although this approach does not say which command should be used when, it does provide a guide. The algorithms developed for three applications are presented. These are: the calibration of a photoelastic material used in the measurement of the foot-ground pressure pattern of a standing person; the measurement of areas within stem disk growth rings; and the detection of surface blemishes of kiwifruit being graded for export.