Sparse Image Coding Using an Asynchronous Spiking Neural Network
This was an attempt to replicate the retina model from [Van Rullen, 01] and to correct the error in reconstruction from this last paper due to the use of the Calderon formula. Looking at the sparsest code, this induced the introduction of correlation-based inhibition to correct this flaw. By applying these corrections, we reinvented an existing algorithm, Matching Pursuit, which was further applied to modelling the primary visual cortex (see (Perrinet,02) and (Perrinet, 03, IEEE))
see slides (PDF) comparing different strategies
get a reprint Perrinet02esann
- Laurent Perrinet, Manuel Samuelides. Visual Strategies for Sparse Spike Coding, URL URL2 . In Actes de Neurosciences et Sciences de l'Ingenieur, L'Agelonde,, 2002 abstract
- Laurent Perrinet, Manuel Samuelides. Sparse Image Coding Using an Asynchronous Spiking Neural Network, URL . In Proceedings of ESANN, pages 313--8. 2002 abstract
All material (c) L. Perrinet. Please check the copyright notice.
Figure 1 Progressive reconstruction of the spiking image in the retina. To illustrate that the visual information is contained in the spike code, we show the theoretical reconstruction of the Lena image using the algorithm presented in the paper. This particular reconstruction on the 256x256 image used a Laplacian pyramid as the linear transform because this transform is invertible and exhibits only little cross-correlation between filters. Results are improved compared to the use of the Calderon frormula (as in [VanRullen, 01]), see Fig. 2 and we recognize the original image after only a few hundreds spikes.