A multi-task neurodynamical model of lateral interactions in V1: visual saliency of colour images

Attention processing mechanisms have been proved to be influenced by the primary visual cortex activity. Attention maps, named saliency maps, are the representation of the most relevant cues that attract attention given a specific stimulus, either synthetic or real images. These saliency maps have been predicted using artificial as well as biologically-inspired models, taking up great results on any kind of stimuli. However, most of these models have been built specifically for visual saliency, a characteristic that denies their biological plausibility of being a unified mechanism of V1 (found to be responsible for processes other than visual saliency). A model defining the primary visual cortex firing rate dynamics has been proposed in Zhaoping Li’s work (1998), in which the lateral connections in V1 are found to extract visual saliency by detecting specific distractors (color, brightness, orientation and scale). Penacchio et al's architecture (2013) was inspired on such mechanism for predicting brightness induction in real images. Our model combines such method with a pipeline similar to the one used in previous work on bottom-up visual saliency from Murray et al. (2011).