A study of real-time and rate scalable image and video compression
Abstract
In this thesis, we address two issues related to image and video compression. The first issue is on how to accelerate the speed of video compression using parallel processing, and the second is the development of a new approach to rate scalable coding of images and video. (1) Several parallel implementations of an MPEG1 video encoder on MIMD high performance systems, in particular the Intel Touchstone Delta and the Intel Paragon, are presented. In our approach, both spatial and temporal parallelism have been exploited. While the Paragon has the computation capacity to compress video sequences at a speed faster than real-time, we found that real-time performance cannot be achieved if the Input/Output (I/O) is not designed properly. We present several schemes, corresponding to different types of data parallelism, for managing the I/O operations and regulating the data flow. Using our algorithm, real-time MPEG compression of ITU-R 601 digital video sequences was achieved. (2) A new wavelet based rate scalable image and video compression algorithm is presented. A new embedded zerotree wavelet (EZW) approach for color image compression that exploits the interdependency between color components in the luminance/chrominance color space is described. The algorithm is known as Color Embedded Zerotree Wavelet (CEZW). The new video compression algorithm uses motion compensation to reduce temporal redundancy. The prediction error frames and the intra-coded frames are encoded using CEZW. To address the error propagation problem inherit to rate scalable video compression, an adaptive motion compensation (AMC) scheme is designed. The rate scalable video compression algorithm is known as Scalable Adaptive Motion Compensation Wavelet (SAMCoW). We show that in addition to providing' a wide range of rate scalability SAMCoW achieves comparable performance to the more traditional hybrid video coders, such as MPEG1 and H.263. Furthermore, SAMCoW allows the data rate to be dynamically changed during decoding, which is very appealing for network oriented applications.
Degree
Ph.D.
Advisors
Delp, Purdue University.
Subject Area
Electrical engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.