Differences between revisions 6 and 18 (spanning 12 versions)
 ⇤ ← Revision 6 as of 2015-02-26 16:25:41 → Size: 1886 Editor: LaurentPerrinet Comment: ← Revision 18 as of 2019-01-19 12:00:59 → ⇥ Size: 4540 Editor: LaurentPerrinet Comment: Deletions are marked like this. Additions are marked like this. Line 4: Line 4: * Browse the [[http://www.eusipco2015.org/content/technical-programme|technical programme]] of the conference * Download the corresponding [[attachment:perrinet15eusipco.pdf|manuscript|&do=get]]. Line 5: Line 7: ##== reproducible science ==## * get the [[attachment:source.zip|open-source scripts]]## * Download [[attachment:results.pdf|supplementary information|&do=get]] (contains many more controls and experiments than in the paper + an overview of what is produced by the [[attachment:source.zip|scripts]]). || [[Publications/PerrinetBednar15|{{attachment:Figures/PerrinetBednar15/FigureModel/figure_model.jpg|Architecture of the model|width=100%,align="left"}}]] <
> ''[[Figures/PerrinetBednar15/FigureModel| Figure 1]]'': '''Edge co-occurrences''' ''(A)'' An example image with the list of extracted edges overlaid. Each edge is represented by a red line segment which represents its position (center of segment), orientation, and scale (length of segment). . ''(B)'' The relationship between a reference edge "A" and another edge "B" can be quantified in terms of the difference between their orientations, ratio of scale, distance between their centers, and difference of azimuth. This is used to compute the chevron map in [[Figures/PerrinetBednar15/FigureChevrons| Figure 2]]. Go back to [[Publications/PerrinetBednar15|manuscript page]]. ||  Titre:: Sparse Coding Of Natural Images Using A Prior On Edge Co-Occurrences  Résumé:: Oriented edges in images commonly occur in co-linear and co-circular  arrangements, obeying the ``good continuation law'' of Gestalt  psychology. The human visual system appears to exploit this  property of images, with contour detection, line completion, and  grouping performance well predicted by such an ``association field''  between edge elements \citep{Field93,Geisler01}. In this paper, we  show that an association field of this type can be used to enhance  the sparse representation of natural images. First, we define the  !SparseLets framework as an efficient representation of images based  on a discrete wavelet transform. Second, we extract second-order  information about edge co-occurrences from a set of images of natural  scenes. Finally, we incorporate this prior information into our  framework and show that it allows for the extraction of features  relevant to natural scenes, like a round shape. This novel approach  points the way to practical computer vision algorithms with   human-like performance. Slides:: [[attachment:perrinet15eusipco_talk.pdf|slides]], [[attachment:perrinet15eusipco_handout.pdf|slides with notes]]  Code:: For more information on Matching Pursuit on natural images, follow [[https://laurentperrinet.github.io/sciblog/posts/2015-05-22-a-hitchhiker-guide-to-matching-pursuit.html|The hitchhiker guide to Matching Pursuit]]. Line 11: Line 36: @INPROCEEDINGS{Perr1508:Sparse,AUTHOR="Laurent Perrinet and James Bednar",TITLE="Sparse Coding Of Natural Images Using A Prior On Edge {Co-Occurences}",BOOKTITLE="European Signal Processing Conference 2015 (EUSIPCO 2015)",ADDRESS="Nice, France",MONTH=aug,YEAR=2015,KEYWORDS="sparse coding; natural scene statistics; sparselets; lateral connections;association field",ABSTRACT="Oriented edges in images of natural scenes tend to be aligned in co-linearor co-circular arrangements, with lines and smooth curves more common thanother possible arrangements of edges (the ``good continuation law'' ofGestalt psychology). The visual system appears to take advantage of thisprior knowledge about natural images, with human contour detection andgrouping performance well predicted by such an ``association field''between edge elements. Geisler et al (2001) have estimated this priorinformation available to the visual system by extracting contours from adatabase of natural images, and showed that these statistics could predictbehavioral data from humans in a line completion task. In this paper, weshow that an association field of this type can be used for the sparserepresentation of natural images." @inproceedings{Perrinet15eusipco,    abstract = {Oriented edges in images of natural scenes tend to be aligned in co-linear or co-circular arrangements, with lines and smooth curves more common than other possible arrangements of edges (the "good continuation law" of Gestalt psychology). The visual system appears to take advantage of this prior knowledge about natural images, with human contour detection and grouping performance well predicted by such an "association field" between edge elements. Geisler et al (2001) have estimated this prior information available to the visual system by extracting contours from a database of natural images, and showed that these statistics could predict behavioral data from humans in a line completion task. In this paper, we show that an association field of this type can be used for the sparse representation of natural images.},    address = {Nice, France},    author = {Perrinet, Laurent U. and Bednar, James A.},    booktitle = {European Signal Processing Conference 2015 (EUSIPCO 2015)},    citeulike-article-id = {13563378},    keywords = {association, bicv-sparse, coding, connections, field, lateral, natural, scene, sparse, sparselets, statistics},    month = aug,    posted-at = {2015-03-27 10:56:57},    priority = {2},    title = {Sparse Coding Of Natural Images Using A Prior On Edge {Co-Occurences}},    year = {2015},url = {http://ieeexplore.ieee.org/document/7362781/}url = {http://dx.doi.org/10.1109/EUSIPCO.2015.7362781},doi = {10.1109/EUSIPCO.2015.7362781}, Line 33: Line 52: Line 43: Line 60: TagYear15 TagBrainScales TagPublicationsInPreparation TagPublicationsInPreparation TagYear15 TagBrainScales TagPublicationsArticles TagTalks

# Sparse coding of natural images using a prior on edge co-occurences

 Figure 1: Edge co-occurrences (A) An example image with the list of extracted edges overlaid. Each edge is represented by a red line segment which represents its position (center of segment), orientation, and scale (length of segment). . (B) The relationship between a reference edge "A" and another edge "B" can be quantified in terms of the difference between their orientations, ratio of scale, distance between their centers, and difference of azimuth. This is used to compute the chevron map in Figure 2. Go back to manuscript page.
Titre
Sparse Coding Of Natural Images Using A Prior On Edge Co-Occurrences
Résumé
Oriented edges in images commonly occur in co-linear and co-circular

arrangements, obeying the good continuation law of Gestalt psychology. The human visual system appears to exploit this property of images, with contour detection, line completion, and grouping performance well predicted by such an association field between edge elements \citep{Field93,Geisler01}. In this paper, we show that an association field of this type can be used to enhance the sparse representation of natural images. First, we define the SparseLets framework as an efficient representation of images based on a discrete wavelet transform. Second, we extract second-order information about edge co-occurrences from a set of images of natural scenes. Finally, we incorporate this prior information into our framework and show that it allows for the extraction of features relevant to natural scenes, like a round shape. This novel approach points the way to practical computer vision algorithms with human-like performance.

Slides
Code

For more information on Matching Pursuit on natural images, follow The hitchhiker guide to Matching Pursuit.

## reference

• Laurent U. Perrinet, James A. Bednar. Sparse Coding Of Natural Images Using A Prior On Edge Co-Occurences, URL URL2 . In European Signal Processing Conference 2015 (EUSIPCO 2015), Nice, France, 2015 abstract.

All material (c) L. Perrinet. Please check the copyright notice.

 This work was supported by European Union project Number FP7-269921, "BrainScales".

welcome:
• Page.execute = 0.003s
• getACL = 0.013s
• i18n_init = 0.022s
• init = 0.022s
• loadLanguage = 0.001s
• load_multi_cfg = 0.018s
• run = 0.170s
• send_page = 0.096s
• send_page_content = 0.096s
• send_page_content|1 = 0.006s
• send_page|1 = 0.010s
• total = 0.193s