An unequal channel protection technique for JPEG2000 encoded images is proposed. The proposed technique takes advantage of the hierarchical organization of the JPEG2000 bit-stream and protects only the most important part of the compressed bit-stream, by means of a rate compatible punctured convolutional code. Both single-quality-layer and multiple-quality-layer compressions are considered. The performance of this technique is evaluated under various noisy channel conditions. Simulation results show that the probability of correct decoding the received information and the quality of the reconstructed image are improved by the proposed technique.