We study an iterative design algorithm that jointly optimizes source and channel codes for image transmission over Rayleigh fading channels. The joint design combines channel-optimized vector quantization (COVQ) for the source code with a family of rate-compatible punctured convolutional codes for the unequal-error-protection (UEP) channel code. We consider both hard-decision and soft-decision decoding for these channel codes. Our objective is to minimize the average end-to-end distortion, averaged over both the source and channel statistics. For a given channel SNR and transmission rate, our joint source and channel code design achieves an optimal allocation of bits between the source and channel coders. This optimal allocation can reduce distortion by up to 7 dB over suboptimal allocations for the source data set considered. Our joint iterative design exhibits a coding gain over COVQ of 1.5 dB for hard-decision decoding and 3 dB for soft-decision decoding.