The `independent components''of natural scenes are edge filters

By Anthony J. Bell, T. Sejnowski. 1997

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

Density Modeling of Images using a Generalized Normalization Transformation (2015, positive): “Variants of divisive normalization have been found to reduce dependencies when applied to natural images or sounds and to produce approximately Gaussian responses (Ruderman, 1994; Schwartz & Simoncelli, 2001; Malo & Laparra, 2010).”
A Unifying Information-Theoretic Framework for Independent Component Analysis (2000, unknown): “Field (1992) has suggested that factorial codes are an e(cid:14)cient coding strategy for visual sensory processing, and ICA applied to natural images yields localized and oriented (cid:12)lters similar to those found in visual cortex (Bell and Sejnowski, 1997).”
Saliency, attention, and visual search: an information theoretic approach. (2009, unknown): “…A number of inﬂuential studies have demonstrated that learning a sparse encoding of natural image statistics may yield a sparse V1-like set of cells (Bell & Sejnowski, 1997; Olshausen & Field, 1996), including those that code for spatiotemporal content (van Hateren & Ruderman, 1998) and color…”
THE CO-INFORMATION LATTICE (2003, unknown): “Radially symmetric distributions have already shown their use in the most successful DCA algorithm to date, Hyv¨ arinen & Hoyer’s subspace-ICA, which, building on [ 4 , 10], accounts for most of the main features of brain area V1 organisation.”
Nonlinear independent component analysis: Existence and uniqueness results (1999, unknown): “ICA can be applied to blind source separation (Jutten and Herault, 1991) as well as to feature extraction (Bell and Sejnowski, 1997; Olshausen and Field, 1996).”
Learning Invariance from Natural Images Inspired by Observations in the Primary Visual Cortex (2012, unknown): “There are many approaches for learning simple-cell properties (Ol-shausen & Field, 1996; Bell & Sejnowski, 1997; van Hateren & van der Schaaf, 1998; Hoyer & Hyv¨arinen, 2000; Falconbridge, Stamps, & Badcock, 2006; Hamker & Wiltschut, 2007; Rehn & Sommer, 2007; Weber & Triesch, 2008; Wiltschut &…”
Learning Low-Level Vision (1999, unknown): “Researchers have related those statistics to properties of the human visual system (Olshausen and Field, 1996; Bell and Sejnowski, 1997; Simoncelli, 1997), or have used statistical characterizations of images to analyse and synthesize realistic textures (Heeger and Bergen, 1995; DeBonet and Viola,…”
Generative Image Modeling Using Spatial LSTMs (2015, unknown): “992 ICA [1, 48] 1.”
Deep Learning in Spiking Neural Networks (2018, unknown): “Representation learning methods, which use neural networks such as autoencoders and sparse coding schemes, learn to discover visual features similar to the receptive field properties found in V1 [161], [162], [163], [164].”
Sparse coding based visual tracking: Review and experimental comparison (2013, unknown): “[21] proposed an appearance model based on the response distribution of basis functions learned using independent component analysis (ICA) [37].”
Natural image statistics and neural representation. (2001, unknown): “Similar results have been obtained using other forms of ICA (Bell & Sejnowski 1997, van Hateren & van der Schaaf 1998, Lewicki & Olshausen 1999), and Hyvärinen & Hoyer (2000) have derived complex cell properties by extending ICA to operate on subspaces.”
Independent Component Analysis (2005, unknown): “ICA has been applied to separation of different speech signals [1]2, analysis of EEG data [6], functional magnetic resonance imaging (fMRI) data [7], image processing [2], and as a model of biological image processing [10].”
Why is Real-World Visual Object Recognition Hard? (2008, unknown): “Although controversial ([7,8]), a popular recent approach in the study of vision is the use of ‘‘natural’’ images [7,9–12], in part because they ostensibly capture the essence of problems encountered in the real world.”
SUN: A Bayesian framework for saliency using natural statistics. (2008, unknown): “…advantages to using ICA features: efﬁcient coding has been proposed as one of the fundamental goals of the visual system (Barlow, 1994), and linear ICA has been shown to generate receptive ﬁelds akin to those found in primary visual cortex (V1) (Bell & Sejnowski, 1997; Olshausen & Field, 1996).”