The dynamics of stabilized images in the retina contributes to support the hypothesis of principal components computation at the level of the thalamus

摘要

A computer model of thalamic pattern completion dynamics is used in this paper to replicate the dynamics of stabilized images in the retina. Stabilized images in the retina are perceived by the subject as a sequence of patterns that appears and suddenly fade away. These patterns are similar to the presented pattern, are contained in it or are related to it. They are hypothesized to correspond to a combination of the principal components of the learned images. These components appear when a tiny portion of the image instead of the full image arrives to the thalamus. This happens because images, when stabilized over the retina, produce a saturation of rods and cones and only a fraction of the original image arrives to the thalamus at each moment. The remaining perceived image is a mental reconstruction (or hallucination) of the image based on the tiny pieces of information coming from the eye to the thalamus.