Abstract : The mammalians retina conveys information by means of spike trains. Though, understanding the way these spike trains represent the stimuli is still a challenging issue, especially when considering their non-determinism. Interestingly, the spike-based code of the retina is binary-like, and thus we considered its study in a signal coding fashion. To do so, we specied a coding scheme based on the mean ring rate of spikes simulated by a realistic model of the mammalians retina. We, then, inverted the generated code to reconstruct the original input. Besides, we established some links between the processing occurring in the retina and state-of-the art methods in pure image coding. Finally, we gave a biologically plausible interpretation for the non-determinism in the spike ring timings.