Low complexity finite-state scalar quantization of image subbands

This paper presents an image coding technique based on the combination of subband decomposition, entropy coding and three different forms of quantization: scalar quantization, finite-state scalar quantization (FSSQ) and predictive FSSQ (PFSSQ). In addition to being fast and requiring no training for quantizer design, the technique has been shown to be competitive with other existing methods. In one of our experiments for example, we encoded the luminance component of the standard image Lena at 0.123 bits per pixel with a PSNR of 30.62 dB.