Adaptive DCT coding of images using entropy-constrained trellis coded quantization

In this paper we develop an adaptive DCT based image coding scheme in which a combination of a perceptually-motivated image model, entropy-constrained trellis coded quantization (ECTCQ) and perceptual error weighting is employed to obtain good subjective performance at low bit rates. The model is used to decompose the image into: (i) strong edge, (ii) slow-intensity variations and (iii) texture components. The perceptually important strong edges are encoded essentially losslessly. The remaining components are encoded using an adaptive DCT in which the transform coefficients are quantized by ECTCQ. The contrast-sensitivity of the human visual system is used for perceptual weighting of the transform coefficients. Simulation results are provided and some comparisons are made.