Second-Order Statistics of Natural Images
Science Direct Working Paper No S1574-034X(04)70540-2
17 Pages Posted: 5 Mar 2018
Date Written: February 2002
Assuming adaptation of the visual cortex to its environment, we analyse the invariance structure found in natural images to explain the selective response of neurons to different modalities.We argue that the invariant structure of images can be formally expressed by dot-product kernels. After learning these kernels using natural images we show, how they can be analyzed by an eigen-sub-space analysis. In addition to localized edge detectors we found neurons that respond to changes in texture and cells that detect edge curvature. The analysis suggests new types of features neurons in primary visual cortex may be selective to.
Keywords: symmetry, edges, texture, curvature
Suggested Citation: Suggested Citation