Off-campus UNL users: To download campus access dissertations, please use the following link to log into our proxy server with your NU ID and password. When you are done browsing please remember to return to this page and log out.
Non-UNL users: Please talk to your librarian about requesting this dissertation through interlibrary loan.
Data compression using adaptive transform coding
Abstract
The data of natural images is not stationary, and the coding complexity of images varies from region to region. How well any particular source coding system works is dependent upon its design assumptions and how well these assumptions match the data. In this dissertation adaptive low-rate source coders are developed. These coders adapt by adjusting the complexity of the coder to match the local coding difficulty of the image. This is accomplished by using a threshold driven maximum distortion criterion to select the specific coder used. The different coders are built using variable blocksized transform techniques, and the threshold criterion selects small transform blocks to code the more difficult regions and larger blocks to code the less complex regions. The different coders are interconnected using a quad tree structure. The algorithm that generates the tree and tests the thresholds is independent of the design of block coders. This allows the system designer to select different coders for the different types of images being coded without having to change the entire coding system. A progressive transmission scheme based upon these coders is developed. A set of example systems are constructed to test the feasibility of these systems of source coders. These systems use scalar quantized and vector quantized transform block coders. Some of the systems are extended to include a new modified block truncation coding scheme. They are used to code both monochromatic and color images with good results to rates as low as 0.3 bits/pel. A theoretical framework is constructed from which the study of these coders can be explored, and an algorithm for selecting the optimal bit allocation for the quantization of transform coefficients is developed. The bit allocation algorithm is more fully developed, and can be used to achieve more accurate bit assignments than the algorithms currently used in the literature. Some upper and lower bounds for the bit-allocation distortion-rate function are developed. An obtainable distortion-rate function is developed for a particular scalar quantizer mixing method that can be used to code transform coefficients at any rate is also presented.
Subject Area
Electrical engineering
Recommended Citation
Rost, Martin Christopher, "Data compression using adaptive transform coding" (1988). ETD collection for University of Nebraska-Lincoln. AAI8904509.
https://digitalcommons.unl.edu/dissertations/AAI8904509