Limits...
Robust object recognition under partial occlusions using NMF.

Soukup D, Bajla I - Comput Intell Neurosci (2008)

Bottom Line: In recent years, nonnegative matrix factorization (NMF) methods of a reduced image data representation attracted the attention of computer vision community.We have analyzed the influence of sparseness on recognition rates (RRs) for various dimensions of subspaces generated for two image databases, ORL face database, and USPS handwritten digit database.One of these metrics also is a novelty we proposed.

View Article: PubMed Central - PubMed

Affiliation: Smart systems division, ARC Seibersdorf research GmbH, 2444 Seibersdorf, Austria. daniel.soukup@arcs.ac.at

ABSTRACT
In recent years, nonnegative matrix factorization (NMF) methods of a reduced image data representation attracted the attention of computer vision community. These methods are considered as a convenient part-based representation of image data for recognition tasks with occluded objects. A novel modification in NMF recognition tasks is proposed which utilizes the matrix sparseness control introduced by Hoyer. We have analyzed the influence of sparseness on recognition rates (RRs) for various dimensions of subspaces generated for two image databases, ORL face database, and USPS handwritten digit database. We have studied the behavior of four types of distances between a projected unknown image object and feature vectors in NMF subspaces generated for training data. One of these metrics also is a novelty we proposed. In the recognition phase, partial occlusions in the test images have been modeled by putting two randomly large, randomly positioned black rectangles into each test image.

No MeSH data available.


Related in: MedlinePlus

Classification results for USPS training image data. (a), (c), (e):  Hoyer'sNmfsc algorithm applied to occluded test images for sW = 0.1, 0.5, 0.9, sH = [ ]. (b), (d), (f): our modified modNMF algorithm applied to occluded test images for the identicalvalues of the sparseness parameters.
© Copyright Policy - open-access
Related In: Results  -  Collection


getmorefigures.php?uid=PMC2396239&req=5

fig9: Classification results for USPS training image data. (a), (c), (e): Hoyer'sNmfsc algorithm applied to occluded test images for sW = 0.1, 0.5, 0.9, sH = [ ]. (b), (d), (f): our modified modNMF algorithm applied to occluded test images for the identicalvalues of the sparseness parameters.

Mentions: In the secondpart of our study, we were interested in a comparison of the RR of Nmfsc andmodNMF, latter one being implemented with Hoyer's sparseness control mechanisms.Of course, since the NMF methodology is intended mainly to generate part-basedsubspace representation of template images, our further interest wasconcentrated only on occluded images. These results, obtained for optimumvalues of sparseness parameter sW, are displayed in Figures 8, 9, and 10. The plotsalso show RR versus subspace dimension r but the columnsnow discriminate the used algorithms (Nmfsc—(a), (c), (e), and modNMF—(b), (d), (f)).The plots with the best recognition results have been chosen.


Robust object recognition under partial occlusions using NMF.

Soukup D, Bajla I - Comput Intell Neurosci (2008)

Classification results for USPS training image data. (a), (c), (e):  Hoyer'sNmfsc algorithm applied to occluded test images for sW = 0.1, 0.5, 0.9, sH = [ ]. (b), (d), (f): our modified modNMF algorithm applied to occluded test images for the identicalvalues of the sparseness parameters.
© Copyright Policy - open-access
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC2396239&req=5

fig9: Classification results for USPS training image data. (a), (c), (e): Hoyer'sNmfsc algorithm applied to occluded test images for sW = 0.1, 0.5, 0.9, sH = [ ]. (b), (d), (f): our modified modNMF algorithm applied to occluded test images for the identicalvalues of the sparseness parameters.
Mentions: In the secondpart of our study, we were interested in a comparison of the RR of Nmfsc andmodNMF, latter one being implemented with Hoyer's sparseness control mechanisms.Of course, since the NMF methodology is intended mainly to generate part-basedsubspace representation of template images, our further interest wasconcentrated only on occluded images. These results, obtained for optimumvalues of sparseness parameter sW, are displayed in Figures 8, 9, and 10. The plotsalso show RR versus subspace dimension r but the columnsnow discriminate the used algorithms (Nmfsc—(a), (c), (e), and modNMF—(b), (d), (f)).The plots with the best recognition results have been chosen.

Bottom Line: In recent years, nonnegative matrix factorization (NMF) methods of a reduced image data representation attracted the attention of computer vision community.We have analyzed the influence of sparseness on recognition rates (RRs) for various dimensions of subspaces generated for two image databases, ORL face database, and USPS handwritten digit database.One of these metrics also is a novelty we proposed.

View Article: PubMed Central - PubMed

Affiliation: Smart systems division, ARC Seibersdorf research GmbH, 2444 Seibersdorf, Austria. daniel.soukup@arcs.ac.at

ABSTRACT
In recent years, nonnegative matrix factorization (NMF) methods of a reduced image data representation attracted the attention of computer vision community. These methods are considered as a convenient part-based representation of image data for recognition tasks with occluded objects. A novel modification in NMF recognition tasks is proposed which utilizes the matrix sparseness control introduced by Hoyer. We have analyzed the influence of sparseness on recognition rates (RRs) for various dimensions of subspaces generated for two image databases, ORL face database, and USPS handwritten digit database. We have studied the behavior of four types of distances between a projected unknown image object and feature vectors in NMF subspaces generated for training data. One of these metrics also is a novelty we proposed. In the recognition phase, partial occlusions in the test images have been modeled by putting two randomly large, randomly positioned black rectangles into each test image.

No MeSH data available.


Related in: MedlinePlus