Image Interpolation Consistent with Neuromorphic Coding Data

In the paper, the image encoding/decoding methods in the context of a neuromorphic model based on known mechanisms of visual perception are under consideration. The input data is a representation of the recorded information in the form of a stream of events, similar to the events of activation of the visual receptors of the retina. The event stream model is selected in the form of a Poisson point process of random retinal counts. Data encoding is carried out within the framework of the concept of receptive fields of the retina. To adequately modelling of neuromorphic coding, a number of known neural mechanisms are formalized, in particular central/lateral inhibition. Decoding of the obtained neuromorphic representations is considered in the context of restoration of spatial contrasts of stimuli, implying modeling of the function of simple cells of the primary visual cortex. The obtained decoding model allows you to restore local boundaries of objects in the image that improves the visual quality of images, for example, in comparison with the quality by standard methods of bilinear, bicubic and other interpolations.