Statistics of Natural Images: Scaling in the Woods

By Daniel L. Ruderman, William Bialek. 1993

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

ConViT: improving vision transformers with soft convolutional inductive biases (2021, neutral): “models more sample-eﬃcient and parameter-eﬃcient (Simoncelli and Olshausen, 2001, Ruderman and Bialek 1994).”
Statistical Edge Detection: Learning and Evaluating Edge Cues (2003, background): “But, as has been widely reported [12], [1], [30], [39], [24], [35], natural images have highly structured statistical properties which typically do not agree with the assumptions made by current edge detectors.”
Escaping the Big Data Paradigm with Compact Transformers (2021, background): “They allow CNNs to leverage natural image statistics and subsequently allow models to have higher sampling efﬁ-ciency [34, 34].”
A two-layer sparse coding model learns simple and complex cell receptive fields and topography from natural images. (2001, background): “First, the visual input is not white noise: The Fourier amplitudes fall off approximately proportional to the inverse of the frequency (see e.g. (Ruderman and Bialek, 1994)).”
Unifying Divergence Minimization and Statistical Inference Via Convex Duality (2006, background): “There the problem of minimizing ‖x‖22 subject to ‖Ax− b‖22 ≤ is studied as a means of improving the stability of the problem Ax = b. – Ruderman and Bialek [18] study a related problem where instead of a quadratic penalty on x the following objective function is minimize −H(x) subject to ‖Ax− b‖22 ≤ ”
Deep Probabilistic Modeling of Natural Images using a Pyramid Decomposition (2016, background): “Probabilistic image modeling is a research area of long tradition that has attracted interest from many different disciplines (Ruderman & Bialek, 1994; Olshausen & Field, 1996; Hyv¨arinen et al., 2009).”
Fast Feature Pyramids for Object Detection (2014, background): “To understand more generally how features behave in resampled images, we turn to the study of natural image statistics [7], [33].”
Slow feature analysis yields a rich repertoire of complex cell properties. (2005, background): “The statistical properties of such images are equivalent to those of natural images up to the second order (Ruderman & Bialek, 1994; Dong & Atick, 1995).”