A comparison of natural-image-based models of simple-cell coding.
Models such as that of Olshausen and Field (OandF, 1997 Vision Research 37 3311-3325) and principal components analysis (PCA) have been used to model simple-cell receptive fields, and to try to elucidate the statistical principles underlying visual coding in area V1. They connect the statistical str...
| Main Authors: | , , |
|---|---|
| Format: | Journal article |
| Language: | English |
| Published: |
2000
|
| _version_ | 1797071169691582464 |
|---|---|
| author | Willmore, B Watters, P Tolhurst, D |
| author_facet | Willmore, B Watters, P Tolhurst, D |
| author_sort | Willmore, B |
| collection | OXFORD |
| description | Models such as that of Olshausen and Field (OandF, 1997 Vision Research 37 3311-3325) and principal components analysis (PCA) have been used to model simple-cell receptive fields, and to try to elucidate the statistical principles underlying visual coding in area V1. They connect the statistical structure of natural images with the statistical structure of the coding used in V1. The OandF model has created particular interest because the basis functions it produces resemble the receptive fields of simple cells. We evaluate these models in terms of their sparseness and dispersal, both of which have been suggested as desirable for efficient visual coding. However, both attributes have been defined ambiguously in the literature, and we have been obliged to formulate specific definitions in order to allow any comparison between models at all. We find that both attributes are strongly affected by any preprocessing (e.g. spectral pseudo-whitening or a logarithmic transformation) which is often applied to images before they are analysed by PCA or the OandF model. We also find that measures of sparseness are affected by the size of the filters--PCA filters with small receptive fields appear sparser than PCA filters with larger spatial extent. Finally, normalisation of the means and variances of filters influences measures of dispersal. It is necessary to control for all of these factors before making any comparisons between different models. Having taken these factors into account, we find that the code produced by the OandF model is somewhat sparser than the code produced by PCA. However, the difference is rather smaller than might have been expected, and a measure of dispersal is required to distinguish clearly between the two models. |
| first_indexed | 2024-03-06T22:49:24Z |
| format | Journal article |
| id | oxford-uuid:5e4d64b1-8f70-4a87-b671-19ed7e184ef9 |
| institution | University of Oxford |
| language | English |
| last_indexed | 2024-03-06T22:49:24Z |
| publishDate | 2000 |
| record_format | dspace |
| spelling | oxford-uuid:5e4d64b1-8f70-4a87-b671-19ed7e184ef92022-03-26T17:39:44ZA comparison of natural-image-based models of simple-cell coding.Journal articlehttp://purl.org/coar/resource_type/c_dcae04bcuuid:5e4d64b1-8f70-4a87-b671-19ed7e184ef9EnglishSymplectic Elements at Oxford2000Willmore, BWatters, PTolhurst, DModels such as that of Olshausen and Field (OandF, 1997 Vision Research 37 3311-3325) and principal components analysis (PCA) have been used to model simple-cell receptive fields, and to try to elucidate the statistical principles underlying visual coding in area V1. They connect the statistical structure of natural images with the statistical structure of the coding used in V1. The OandF model has created particular interest because the basis functions it produces resemble the receptive fields of simple cells. We evaluate these models in terms of their sparseness and dispersal, both of which have been suggested as desirable for efficient visual coding. However, both attributes have been defined ambiguously in the literature, and we have been obliged to formulate specific definitions in order to allow any comparison between models at all. We find that both attributes are strongly affected by any preprocessing (e.g. spectral pseudo-whitening or a logarithmic transformation) which is often applied to images before they are analysed by PCA or the OandF model. We also find that measures of sparseness are affected by the size of the filters--PCA filters with small receptive fields appear sparser than PCA filters with larger spatial extent. Finally, normalisation of the means and variances of filters influences measures of dispersal. It is necessary to control for all of these factors before making any comparisons between different models. Having taken these factors into account, we find that the code produced by the OandF model is somewhat sparser than the code produced by PCA. However, the difference is rather smaller than might have been expected, and a measure of dispersal is required to distinguish clearly between the two models. |
| spellingShingle | Willmore, B Watters, P Tolhurst, D A comparison of natural-image-based models of simple-cell coding. |
| title | A comparison of natural-image-based models of simple-cell coding. |
| title_full | A comparison of natural-image-based models of simple-cell coding. |
| title_fullStr | A comparison of natural-image-based models of simple-cell coding. |
| title_full_unstemmed | A comparison of natural-image-based models of simple-cell coding. |
| title_short | A comparison of natural-image-based models of simple-cell coding. |
| title_sort | comparison of natural image based models of simple cell coding |
| work_keys_str_mv | AT willmoreb acomparisonofnaturalimagebasedmodelsofsimplecellcoding AT wattersp acomparisonofnaturalimagebasedmodelsofsimplecellcoding AT tolhurstd acomparisonofnaturalimagebasedmodelsofsimplecellcoding AT willmoreb comparisonofnaturalimagebasedmodelsofsimplecellcoding AT wattersp comparisonofnaturalimagebasedmodelsofsimplecellcoding AT tolhurstd comparisonofnaturalimagebasedmodelsofsimplecellcoding |