A joint source-channel coding system for image communication over an additive white Gaussian noie channel is presented. It employs vector quantization based hybrid digital-analog modulation techniques with bandwidth compression and expansion for transmitting and reconstructing the wavelet coefficients of an image. The main advantage of the proposed system is that it achieves good performance at the design channel signal-to-noise ratio (CSNR), while still maintaining a "graceful improvement" characteristic at higher CSNRs. Comparisons are made with two purely digital systems and two purely analog systems. Simulation shows that the proposed system is superior to the other investigated systems for a wide range of CSNRs.