Machine Vision has been extensively applied in industry. This thesis project, which originated with a local food processor, applies a vision system to inspection of packages for cosmetric errors. The basic elements and theory of the machine vision system are introduced, and some image processing techniques, such as histogram analysis, thresholding, and SRI algorithm, are utilized in this thesis. Computer programs written in C and Pascal are described. Hardware setup and computer interface, such as RS-232 serial interface, parallel digital I/O interface, conveyor control, and incremental shaft encoder, are described. Test results are presented and discussed. / Graduation date: 1991

Graduation date: 1991

Applications of imaging and image processing become a part of our daily life and find their crucial way in real-world areas. Accordingly, the corresponding techniques get more and more complicated. Many tasks are recognizable for a image processing chain, such as, filtering, color balancing, enhancement, segmentation, and post processing. Generally speaking, all of the image processing techniques need a control parameter setting. The better these parameters are set the better results can be achieved. Usually, these parameters are real numbers so search space is really large and brute-force searching is impossible or at least very time consuming. Therefore, the optimal setting of the parameters is an essential requirement to obtain desirable results. Obviously, we are faced with an optimization problem, which its complexity depends on the number of the parameters to be optimized and correlation among them. By reviewing the optimization methods, it can be understood that metaheuristic algorithms are the best candidates for these kind of problems. Metaheuristic algorithms are iterative approaches which can search very complex large spaces to come up with an optimal or close to optimal solution(s). They are able to solve black-box global optimization problems which are not solvable by classic mathematical methods. The first part of this thesis optimizes the control parameters for an eye-illusion, image enhancement, and image thresholding tasks by using an interactive evolutionary optimization approach. Eye illusion and image enhancement are subjective human perception-based issues, so, there is no proposed analytical fitness function for them. Their optimization is only possible through interactive methods. The second part is about setting of active contour (snake) parameters. The performance of active contours (snakes) is sensitive to its eight correlated control parameters which makes the parameter setting problem complex to solve. In this work, wehave tried to set the parameters to their optimal values by using a sample segmented image provided by an expert. As our case studies, we have used breast ultrasound, prostate ultrasound, and lung X-ray medical images. The proposed schemes are general enough to be investigated with other optimization methods and also image processing tasks. The achieved experimental results are promising for both directions, namely, interactive-based image processing and sample-based medical image segmentation. / UOIT

Most of the existing research on mathematical morphology is restricted to the deterministic case. This thesis addresses the void in the results on the stochastic properties of morphological filters. The primary results include analysis of the stochastic properties of morphological operations, such as dilation, erosion, closing and opening. Two unbiased morphological filters are introduced and a quantitative description of the probability distribution function of morphological operations on independent, identically distributed random signals is obtained. Design of an optimal morphological filter in the sense of a criterion proposed here is also discussed. A brief, but systematic description of the definitions and properties of deterministic morphological operations on sets is presented to establish the necessary background for the analysis of the filter stochastic properties. / Graduation date: 1992

Facial feature point detection is a key issue in facial image processing. One main challenge of facial feature point detection is the variation of facial structures due to expressions. This thesis aims to explore more accurate and robust facial feature point detection algorithms, which can facilitate the research on facial image processing, in particular the facial expression analysis. This thesis introduces a facial feature point detection system, where the Multilinear Principal Component Analysis is applied to extract the highly descriptive features of facial feature points. In addition, to improve the accuracy and efficiency of the system, a skin color based face detection algorithm is studied. The experiment results have indicated that this system is effective in detecting 20 facial feature points in frontal faces with different expressions. This system has also achieved a higher accuracy during the comparison with the state-of-the-art, BoRMaN.

Facial feature point detection is a key issue in facial image processing. One main challenge of facial feature point detection is the variation of facial structures due to expressions. This thesis aims to explore more accurate and robust facial feature point detection algorithms, which can facilitate the research on facial image processing, in particular the facial expression analysis. This thesis introduces a facial feature point detection system, where the Multilinear Principal Component Analysis is applied to extract the highly descriptive features of facial feature points. In addition, to improve the accuracy and efficiency of the system, a skin color based face detection algorithm is studied. The experiment results have indicated that this system is effective in detecting 20 facial feature points in frontal faces with different expressions. This system has also achieved a higher accuracy during the comparison with the state-of-the-art, BoRMaN.

Tone mapping is a computational task of significance in the context of displaying high dynamic range images on low dynamic range devices. While a number of tone mapping algorithms have been proposed and are in common use, there is no single operator that yields optimal results under all conditions. Moreover, obtaining satisfactory mappings often requires the manual tweaking of parameters. This thesis proposes interactive evolution as a computational tool for tone mapping. An evolution strategy that blends the results from several tone mapping operators while at the same time adapting their parameters is proposed. As well, the results are adapted such that such that approximately uniform perceptual distances between offspring candidate solutions and the parent are ensured. The introduction of a perceptually based step size adaptation technique enhances the control of the variability between newly generated offspring, when compared to parameter space step size adaptation.

Traditional signal processing systems, based on linear modeling principles, face a stifling pressure to meet present-day demands caused by the deluge of data generated, transmitted and processed across the globe. Fortunately, recent advances have resulted in the emergence of more sophisticated, nonlinear signal models. Such nonlinear models have inspired fundamental changes in which information processing systems are designed and analyzed. For example, the sparse signal model serves as the basis for Compressive Sensing (CS), an exciting new framework for signal acquisition. In this thesis, we advocate a geometry-based approach for nonlinear modeling of signal ensembles. We make the guiding assumption that the signal class of interest forms a nonlinear low-dimensional manifold belonging to the high-dimensional signal space. A host of traditional data models can be essentially interpreted as specific instances of such manifolds. Therefore, our proposed geometric approach provides a common framework that can unify, analyze, and significantly extend the scope of nonlinear models for information acquisition and processing. We demonstrate that the geometric approach enables new algorithms and analysis for a number of signal processing applications. Our specific contributions include: (i) new convex formulations and algorithms for the design of linear systems for data acquisition, compression, and classification; (ii) a general algorithm for reconstruction, deconvolution, and denoising of signals, images, and matrix-valued data; (iii) efficient methods for inference from a small number of linear signal samples, without ever resorting to reconstruction; and, (iv) new signal and image representations for robust modeling and processing of large-scale data ensembles.

This thesis focuses on spatial resolution enhancement of magnetic resonance imaging (MRI). In particular, it addresses methods of performing such enhancement in the Fourier domain. After a brief review of Fourier theory, the thesis reviews the physics of the MRI acquisition process in order to introduce a mathematical model of the measured data. This model is later used to develop and analyze methods for resolution enhancement, or "super-resolution'', in MRI. We then examine strategies of performing super-resolution MRI (SRMRI). We begin by exploring strategies that use multiple data sets produced by spatial translations of the object being imaged, to add new information to the reconstruction process. This represents a more detailed mathematical examination of the author's Master's work at the University of Calgary. Using our model of the measured data developed earlier in the thesis, we describe how the acquisition strategy determines the efficacy of the SRMRI process that employs multiple data sets. The author then explores the self-similarity properties of MRI data in the Fourier domain as a means of performing spatial resolution enhancement. To this end, a fractal-based method over (complex-valued) Fourier Transforms of functions with compact spatial support, derived from a fractal transform in the spatial domain, is explored. It is shown that this method of "Iterated Fourier Transform Systems" (IFTS) can be tailored to perform frequency extrapolation, hence spatial resolution enhancement. The IFTS method, however, is limited in scope, as it assumes that a spatial function f(x) may be approximated by linear combinations of spatially-contracted and range-modified copies of the entire function. In order to improve the approximation, we borrow from traditional fractal image coding in the spatial domain, where subblocks of an image are approximated by other subblocks, and employ such a block-based strategy in the Fourier domain. An examination of the statistical properties of subblock approximation errors shows that, in general, Fourier data can be locally self-similar. Furthermore, we show that such a block-based self-similarity method is actually equivalent to a special case of the auto-regressive moving average (ARMA) modeling method. The thesis concludes with a chapter on possible future research directions in SRMRI.

No description available.