Trellis-based Image Communication over Noisy Channels
Benjamin Belzer, Washington State University School of EECS
Objective
The objective of this research is to develop joint source channel coding (JSCC)
techniques for trellis-based source and channel coding of images and video. These
techniques enable bandwidth efficient transmission of images and video over wireless channels.
Achievements
- Transmission of good quality images over AWGN channels with BPSK bit error rates of 10-1
- Introduction of Symmetric Trellis Coded Vector Quantization (S-TCVQ), a fixed
wordlength vector quantizer well suited for image transmission over noisy channels.
At low bit rates (1 bit/sample) on the Laplacian source (often used to model image
transform coefficients), S-TCVQ matches the performance of trellis-based scalar vector
quantization, the best fixed-wordlength quantizer thus far reported, but requires
less computational complexity.
Example of trellis-based joint source channel coding of images. 512x512 image coded
at 80 kilobytes (compression ratio 26:1), including channel coding bits. AWGN channel
with 10-1 BER, BPSK modulation. Source coding with S-TCVQ of wavelet transform
coefficients, channel coding with soft-decision convolutional code. The source and
channel codes use the same trellis processor.
Click here for an enlarged version of the image
Current Work
- Optimization of S-TCVQ quantization performance through improved set partitioning techniques.
- Optimization of the mapping between source and channel symbols in trellis-based JSCC systems.
- Application of trellis-based JSCC to video communication systems employing interframe coding.
- Development of improved wireless channel models, and application of these models
to image/video communication systems.