Una implementación computacional de un modelo de atención visual Bottom-up aplicado a escenas naturales

The bottom-up visual attention model proposed by Itti et al. 2000 [1], has been a popular model since it exhibits certain neurobiological evidence of primates’ vision. This work complements the computational model of this phenomenon using a neural network with realistic dynamics. This approximation is based on several topographical maps representing the objects saliency that construct a general representation (saliency map), which is the input for a dynamic neural network, whose local and global collaborative and competitive interactions converge to the main particularities (objects) presented by the visual scene as well.