Thesis (Ph. D.)--University of Washington, 2000. / Vita. Includes bibliographical references (p. 110-119).

The adaptation of an IIR filter is a very difficult problem due to its non-quadratic performance surface and potential instability. Conventional adaptive IIR algorithms suffer from potential instability problems and a high cost for stability monitoring. Therefore, there is much interest in adaptive IIR filters based on alternative algorithms. Genetic algorithms are a family of search algorithms based on natural selection and genetics. They have been successfully used in many different areas. Genetic algorithms applied to the adaptation of IIR filtering problems are studied in this thesis, and show that the genetic algorithm approach has a number of advantages over conventional gradient algorithms, particularly, for the adaptation of high order adaptive IIR filters, IIR filters with poles close to the unit circle and IIR filters with multi-modal error surfaces. The conventional gradient algorithms have difficulty solving these problems. Coefficient results are presented for various orders of IIR filters in this thesis. In the computer simulations presented in this thesis, the direct, cascade, parallel and lattice form IIR filter structures have been used and compared. The lattice form IIR filter structure shows its superiority over the cascade and parallel form IIR filter structures in terms of its mean square error convergence performance.

As the costs of digital signal processing and memory hardware are decreasing each year compared to those of transmission, it is increasingly economical to apply sophisticated source encoding techniques to reduce the transmission time for facsimile documents. With this intent, information lossy encoding schemes have been investigated in which the encoder is divided into two stages. Firstly, preprocessing, which removes redundant information from the original documents, and secondly, actual encoding of the preprocessed documents.

By using transform coding, image transmission rates as low as 0.5 bit/pel can be achieved. Generally, the bit rate reduction is achieved by allocating fewer bits to low energy high order coefficients, However, to ensure reasonably good picture quality, a large number of bits has to be allocated to high energy dc coefficients for both fine quantization and good channel error immunity, A technique has been developed that, in some cases, allows the de coefficients to be estimated at the receiver, thus eliminating a major source of difficulty with respect to channel errors.

The aim of this study is to devise detectors for digital satellite modems, that have tolerances to additive white Gaussian noise which are as close as possible to that for optimal detection, at a fraction of the equipment complexity required for optimal detection. Computer simulation tests and theoretical analyses are used to compare the proposed detectors.

No description available.

No description available.

Many practical and experimental systems employing image processing techniques have been built by other workers for various applications. Most of these systems are computer-based and very few operate in a real time environment. The objective of this work is to build a microprocessor-based system for video image processing. The system is used in conjunction with an on-line TV camera and processing is carried out in real time. The enormous storage requirement of digitized TV signals and the real time constraint suggest that some simplification of the data must take place prior to any viable processing. Data reduction is attained through the representation of objects by their edges, an approach often adopted for feature extraction in pattern recognition systems. A new technique for edge detection by applying comparison criteria to differentials at adjacent pixels of the video image is developed and implemented as a preprocessing hardware unit. A circuit for the generation of the co-ordinates of edge points is constructed to free the processing computer of this task, allowing it more time for on-line analysis of video signals. Besides the edge detector and co-ordinate generator the hardware built consists of a microprocessor system based on a Texas Instruments T.US 9900 device, a first-in-first-out buffer store and interface circuitry to a TV camera and display devices. All hardware modules and their power supplies are assembled in one unit to provide a standalone instrument. The problem chosen for investigation is analysis of motion in a visual scene. Aspects of motion studied concern the tracking of moving objects with simple geometric shapes and description of their motion. More emphasis is paid to the analysis of human eye movements and measurement of its point-of-regard which has many practical applications in the fields of physiology and psychology. This study provides a basis for the design of a processing unit attached to an oculometer to replace bulky minicomputer-based eye motion analysis systems. Programs are written for storage, analysis and display of results in real time.

The thesis investigates various aspects of convolutional code (CC) theory, design and decoding. A relatively novel approach to CC theory has been developed. It is based on the concept of convolution and the properties of the encoder (a linear sequential circuit). Some new concepts and results were obtained. The complexity of the trellis diagram of a normal CC encoder is related to code parameters. The associated constrained trellis is obtained by deleting states and transitions, according to a certain criterion, while its complexity is linked to code parameters. Finally the simplified trellis evolves from the constrained one by permitting transitions of length more than one time unit.

This thesis is concerned with the problem of motion estimation and segmentation, mainly related to planar motion in the image plane. The emphasis is placed on several important issues, namely: the study of different motion models and their performance, the benefits resulting from the use of contextual information, the application of multiresolution strategies, and the use of Robust methods and confidence measures. The thesis investigates the application of global motion models, in particular the affine model, in different estimation and segmentation approaches. It is shown that the use of such models, which globally constrain the estimate, results in improved accuracy and robustness. Robust techniques, which can cope with outliers often present when larger data sets are used, are adopted and tested here. The performance is further improved by the use of confidence measures, and of contextual information such as intensity edges or moving feature information. Two broad classes of approach are developed and investigated. The first one is based on the theory of Markov Random Fields. Novel elements in this approach include the introduction of a complex motion model - capable of describing translation, rotation and change of scale - and confidence factors describing the reliability of the data. The application of the Supercoupling approach for multiresolution optimisation speeds up convergence and further improves the quality of the estimate. The second class of algorithms is based on the Hough Transform. An in-depth investigation of the behaviour of the standard Hough Transform is conducted. This leads to the adoption of a robust statistics method providing a better estimate accuracy, better motion segmentation and guaranteed convergence. The use of multiresolution representation in the image plane, in addition to multiresolution in the parameter space, brings the advantage of robust and fast convergence even for large displacements. An important contribution of the research is the evaluation of different kernel functions from the point of view of robustness to noise and change in illumination conditions. Two algorithms from this group have been developed. The first one processes an entire image and provides parallel motion segmentation and estimation. The other is used as a local and robust method for the estimation of optic flow, with the ability to detect multimodal motions. A comparative study with other state-of-the-art methods is conducted, and the results are strongly in favour of the new algorithms. In summary, all stages of motion estimation and segmentation have been investigated. At the low-level, a robust algorithm for optic flow estimation has been developed. It can cope with multiple moving objects, and detects motion boundaries and occluded/uncovered regions. The spatial coherence of motion is enforced here very strongly, resulting in an accurate estimate and reliable confidence measures. This low-level estimate may be globally interpreted, together with other clues and a priori knowledge of the world using a multi-scale Markov Random Field approach. Alternatively, motion estimation and segmentation may be performed in parallel globally using the Robust Hough Transform approach. At this stage meaningful objects can be segmented, thus providing a high-level description of the scene.