What does the retina know about natural scenes?
Neural Computation
A Model for Neuronal Signal Representation by Stimulus-Dependent Receptive Fields
ICANN '09 Proceedings of the 19th International Conference on Artificial Neural Networks: Part I
| Hi-index | 0.00 |
A model for image coding based on Gabor-like stimulusdependent receptive fields has been recently introduced, where the parameters of the coding functions are obtained from the Fourier transform of the input. Here we extend suchmodel, by showing that it can also be based on center-surround structures, such as found in the retina and the lateral geniculate nucleus. Consistent with the interpretation of these early visual stages as providing a decorrelated signal to the cortex, we propose center-surround stimulus-dependent receptive fields which yield whitened representations of natural inputs. Our model receptive fields are found to replicate properties of the mammalian center-surround structures.