Low bit-rate transform and wavelet image coding.

Abstract

In this dissertation, adaptive wavelet and transform coding techniques are presented for low bit-rate monochrome and color image coding. The proposed encoders are based on trellis coded quantization. Trellis coded quantization (TCQ) is an effective scheme for quantizing memoryless sources with low to moderate complexity. The TCQ approach to data compression has led to some of the most effective source codes found to date for memoryless sources. For the transform coder, TCQ is used to encode transform coefficients resulting from applying a 16 x 16 discrete cosine transform (DCT) to 8-bit gray level and 24-bit color images. For the color images, the red, green, and the blue planes were transformed into NTSC transmission primaries (Y, I, and Q) before the DCT is applied. Both fixed-rate and entropy-constrained systems are considered. The discrete wavelet transform has recently emerged as a powerful technique for decomposing images into various multi-resolution approximations. We investigate the use of entropy-constrained trellis coded quantization for encoding the wavelet coefficients of both monochrome and color images. The lowest resolution sub-image is encoded using a 4 x 4 2-D DCT encoder. An integer programming algorithm is employed to allocate the available bit-rate optimally among the subbands. The objective performance results of our wavelet and transform coders are comparable to or surpass all previous results reported in the literature. The subjective quality of the encoded images is also excellent. In particular, the encoded monochrome images at 0.5 bits/pixel (a compression ratio of 16:1) obtained using our adaptive wavelet coder is almost indistinguishable from the original even when viewed on a high-resolution monitor.