Robust Source Coding of Images With Predictive Trellis - Coded Quantization

摘要

The ability to transmit images over narrow bandwidth noisy channels has becomedesirable for many applications. Image integrity and timely transmission are imperative in many scenarios. The traditional method of image transmission requires a multistage process. The first stage is source coding, or the removal of redundancy, to compress the image for the narrow bandwidth channel. The second stage is channel coding, or the adding of redundant characters to protect the information from noise. This report pursues a method to perform robust source coding, providing both compression and noise mitigation. Specifically, Predictive Trellis Coding Quantization (PTCQ) incorporating various types of prediction filters is investigated. Trellis Coded Quanitization (TCQ) implies using an expanded set of quantization levels and the Viterbi algorithm to determine the minimal distortion path through a trellis, whose structure allows for low bit rate encoding. The prediction filter supplies correlation of the prediction differences and thereby provides the protection from noise at the decoder. PTCQ combines TCQ's encoding efficiency with predictive coding compression merits. Linear and nonlinear filter performance within the PTCQ scheme is shown under various channel conditions. Findings show that nonlinear filter implementation provides the highest noise immunity of those tested. The resulting algorithm is implementable in near real-time, allowing for the fast, efficient transmission of images over noisy channels.