The principal components of natural images

@article{Hancock1992ThePC,
  title={The principal components of natural images},
  author={Peter J. B. Hancock and Roland J. Baddeley and Leslie S. Smith},
  journal={Network: Computation In Neural Systems},
  year={1992},
  volume={3},
  pages={61-70}
}
A neural net was used to analyse samples of natural images and text. For the natural images, components resemble derivatives of Gaussian operators, similar to those found in visual cortex and inferred from psychophysics. While the results from natural images do not depend on scale, those from text images are highly scale dependent. Convolution of one of the text components with an original image shows that it is sensitive to inter-word gaps. 

Figures from this paper

Localized principal components of natural images-an analytic solution
The structure of receptive fields in the visual cortex is believed to be shaped by unsupervised learning. A simple variant of unsupervised learning is the extraction of principal components. In this
The principal components of natural images revisited
  • G. Heidemann
  • Environmental Science
    IEEE Transactions on Pattern Analysis and Machine Intelligence
  • 2006
TLDR
A horizontal and vertical typology of PCs is found which leads to the identification of groups of basis functions for steerable bandpass filters and the contribution of spatio-chromatic structure to the total variance can be quantified for selected spatial frequencies.
A statistical analysis of natural images matches psychophysically derived orientation tuning curves
  • R. Baddeley, P. Hancock
  • Psychology
    Proceedings of the Royal Society of London. Series B: Biological Sciences
  • 1991
TLDR
A neural net method is used to extract principal components from real-world images and two of the components are ‘bar-detectors’, which are similar to that suggested by Foster & Ward to account for brief-exposure psychophysical data.
Natural image statistics and visual processing
TLDR
The hypothesis that evolution has optimised the biological visual system to process the images that it commonly encounters i.e., natural images is followed and the structure of natural images with statistical methods is investigated to gain a better understanding of the visual processing in biological systems.
The statistics of natural images
Recently there has been a resurgence of interest in the properties of natural images. Their statistics are important not only in image compression but also for the study of sensory processing in
Natural Image Statistics and Eecient Coding
TLDR
It is suggested that a good objective for an eecient coding of natural Scenes is to maximize the sparseness of the representation, and it is shown that a network that learns sparse codes of natural scenes succeeds in developing localized, oriented, bandpass receptive elds similar to those in the primate striate cortex.
A Survey of Image Statistics Relevant to Computer Graphics
TLDR
The state of the art in image statistics is reviewed and existing and potential applications within computer graphics and related areas are discussed.
A Survey of Image Statistics in Computer Graphics
TLDR
The state of the art in image statistics is reviewed and existing and potential applications within computer graphics and related areas are discussed.
Natural image statistics and efficient coding.
TLDR
It is suggested that a good objective for an efficient coding of natural Scenes is to maximize the sparseness of the representation, and it is shown that a network that learns sparse codes of natural scenes succeeds in developing localized, oriented, bandpass receptive fields similar to those in the mammalian striate cortex.
Regularization operators for natural images based on nonlinear perception models
TLDR
This work proposes the use of biological behavior to build regularization operators as an alternative to simple statistical models and suggests that if the penalty operator takes these additional features in natural images into account, it will be more robust and the choice of the regularization parameter is less critical.
...
...

References

SHOWING 1-10 OF 10 REFERENCES
A statistical analysis of natural images matches psychophysically derived orientation tuning curves
  • R. Baddeley, P. Hancock
  • Psychology
    Proceedings of the Royal Society of London. Series B: Biological Sciences
  • 1991
TLDR
A neural net method is used to extract principal components from real-world images and two of the components are ‘bar-detectors’, which are similar to that suggested by Foster & Ward to account for brief-exposure psychophysical data.
Self-organization in a perceptual network
TLDR
It is shown that even a single developing cell of a layered network exhibits a remarkable set of optimization properties that are closely related to issues in statistics, theoretical physics, adaptive signal processing, the formation of knowledge representation in artificial intelligence, and information theory.
Introduction to the theory of neural computation
TLDR
This book is a detailed, logically-developed treatment that covers the theory and uses of collective computational networks, including associative memory, feed forward networks, and unsupervised learning.
Simplified neuron model as a principal component analyzer
  • E. Oja
  • Biology
    Journal of mathematical biology
  • 1982
A simple linear neuron model with constrained Hebbian-type synaptic modification is analyzed and a new class of unconstrained learning rules is derived. It is shown that the model neuron tends to
Asymmetries in oriented-line detection indicate two orthogonal filters in early vision
  • D. Foster, P. Ward
  • Biology
    Proceedings of the Royal Society of London. Series B: Biological Sciences
  • 1991
Visual detection of a line target differing in orientation from a background of lines may be achieved speedily and effortlessly. Such performance is assumed to occur early in vision and to involve
Optimal unsupervised learning in a single-layer linear feedforward neural network
Neuronal Operations in the Visual Cortex
TLDR
The Visual System of Cat and Monkey Compared, a comparison of the Basic Layout of the Visual System in Cat, Owl Monkey, and Rhesus Monkey, reveals a similar structure to the Retinotopic Organization in the Primary Complex.
Analysis of Linsker's application of Hebbian rules to linear networks
TLDR
The authors analyse the dynamics of the learning rule of a Hebb-type synaptic plasticity rule in a feedforward linear network in terms of the eigenvectors of this matrix, which represent independently evolving weight structures.
From basic network principles to neural architecture
Learning receptive elds
  • IEEE International Conference on Neural Networks, (San Diego 1987), IEEE, New York,
  • 1987