One of the primary design goals of wireless multi-media communication systems is to optimize the quality of user experience. This metric is often quantified using the end-to-end distortion between the original source at the encoder and the reconstructed version of it at the destination. Considering this performance metric, optimal multimedia transmission schemes using joint source-channel coding (JSCC) for fading cognitive radio networks are studied in this paper. Different underlay transmission protocols, based on single- or multi-user receiver capabilities, are compared in terms of their average distortion performance, using an outage analysis. It is demonstrated that JSCC provides significant performance gains over goodput-maximizing schemes that rely on the optimality of source-channel separation, and that, for the scenarios evaluated, one of the proposed protocols with JSCC achieves average distortion performance very close to a lower bound obtained by ignoring the primary interference.