A wavelet-based robust image coding technique for image transmission over noisy channels is presented here. The system combines classification and optimal bit allocation along with robust quantization to produce an inherent error-resistant bitstream without channel coding. In order to improve coding efficiency, an iterative algorithm is developed for implementing optimal bit allocation involving both rate distortion and channel optimization. Also, a modified channel optimized uniform trellis coded quantization (TCQ) with a low computational complexity, yet an effective performance, is proposed and used to provide robust quantization. The system achieves both efficient image compression and robust image transmission over noisy channels. Computer simulations indicate that the proposed scheme is effective, even under mismatched channel conditions, and the performance is competitive with other robust image compression techniques.