Still and Moving Image Compression Systems Using Multiscale Techniques

Citation

Lee, Ming-Chieh (1994) Still and Moving Image Compression Systems Using Multiscale Techniques. Dissertation (Ph.D.), California Institute of Technology. doi:10.7907/02w7-rg28. https://resolver.caltech.edu/CaltechETD:etd-12042007-132203

Abstract

Multi-scale techniques have been popular methods for image and video compression. The basic idea behind these techniques is to decompose the original signals into several components of different scales, however the scale is defined, of different sizes. One then applies appropriate encoding strategies to different components to achieve compression by taking advantage of various properties. In this thesis, we review and present several new schemes of multi-scale techniques using linear and nonlinear systems. Linear techniques, which use linear filters for decomposition, have been thoroughly investigated and widely applied because of their simplicity in implementation and analysis. Herein we describe how to appropriately combine these techniques in order to process the signals more efficiently and advantageously; moreover, the picture quality of the quantized images can be improved. The sub-band coding technique is used as the basis for these combinations. As for the nonlinear technique, we would like to take advantage of the nonlinear features of images (such as edges) in compression to achieve compressing and enhancing of the images. Herein we design several nonlinear multi-resolution systems, using various nonlinear filters, to decompose the signals in a proper form. We show that, in terms of rate-distortion performance, where mean squared error is used as the distortion criterion, these schemes are close to, or even better than, JPEG standard, whereas the encoding and, especially, decoding complexity is lower than that of JPEG. We can obtain much better image quality (in the perceptual sense), however, by applying suitable simple and fast lossless compression schemes to sub-images. Simulation results are demonstrated to show the advantages and feasibility of the proposed scheme. In summary, we mention the current status and future trends of compression technologies.

Thesis Files