Another look at the retina as an image dithered scalar quantizer

We explore, in this paper, the behavior of the mammalians retina considered as an analog-to-digital converter for the incoming light stimuli. This work extends our previous effort towards combining results in neurosciences with image processing techniques. We base our study on a biologically realistic model that reproduces the neural code as generated by the retina. The neural code, that we consider here, consists of non-deterministic temporal sequences of uniformly shaped electrical impulses, also termed as spikes. We describe, starting from this spike-based code, a dynamic quantization scheme that relies on the so-called rate coding hypothesis. We, then, propose a possible decoding procedure. This yields an original coding/decoding system which evolves dynamically from coarse to fine, and from uniform to non-uniform. Furthermore, we emit a possible interpretation for the non-determinism observed in the spike timings. In order to do this, we implement a three-staged processing system mapping the anatomical architecture of the retina. We, then, model the retinal noise by a dither signal which permits us to define the retina behavior as a non-subtractive dithered quantizer. The coding/decoding system, that we propose, offers several interesting features as time scalability as well as reconstruction error whitening and de-correlation from the input stimuli.