TY - GEN
T1 - Learning dictionaries for local sparse coding in image classification
AU - Thiagarajan, Jayaraman J.
AU - Spanias, Andreas
PY - 2011
Y1 - 2011
N2 - Low dimensional embedding of data samples lying on a manifold can be performed using locally linear modeling. By incorporating suitable locality constraints, sparse coding can be adapted to modeling local regions of a manifold. This has been coupled with the spatial pyramid matching algorithm to achieve state-of-the-art performance in object recognition. In this paper, we propose an algorithm to learn dictionaries for computing local sparse codes of descriptors extracted from image patches. The algorithm iterates between a local sparse coding step and an update step that searches for a better dictionary. Evaluation of the local sparse code for a data sample is simplified by first estimating its neighbors using the proposed distance metric and then computing the minimum ℓ 1 solution using only the neighbors. The proposed dictionary update ensures that the neighborhood of a training sample is not changed from one iteration to the next. Simulation results demonstrate that the sparse codes computed using the proposed dictionary achieve improved classification accuracies when compared to using a K-means dictionary with standard image datasets.
AB - Low dimensional embedding of data samples lying on a manifold can be performed using locally linear modeling. By incorporating suitable locality constraints, sparse coding can be adapted to modeling local regions of a manifold. This has been coupled with the spatial pyramid matching algorithm to achieve state-of-the-art performance in object recognition. In this paper, we propose an algorithm to learn dictionaries for computing local sparse codes of descriptors extracted from image patches. The algorithm iterates between a local sparse coding step and an update step that searches for a better dictionary. Evaluation of the local sparse code for a data sample is simplified by first estimating its neighbors using the proposed distance metric and then computing the minimum ℓ 1 solution using only the neighbors. The proposed dictionary update ensures that the neighborhood of a training sample is not changed from one iteration to the next. Simulation results demonstrate that the sparse codes computed using the proposed dictionary achieve improved classification accuracies when compared to using a K-means dictionary with standard image datasets.
KW - Local sparse codes
KW - dictionary learning
KW - linear classifiers
KW - sparse representations
UR - http://www.scopus.com/inward/record.url?scp=84861314060&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84861314060&partnerID=8YFLogxK
U2 - 10.1109/ACSSC.2011.6190379
DO - 10.1109/ACSSC.2011.6190379
M3 - Conference contribution
AN - SCOPUS:84861314060
SN - 9781467303231
T3 - Conference Record - Asilomar Conference on Signals, Systems and Computers
SP - 2014
EP - 2018
BT - Conference Record of the 45th Asilomar Conference on Signals, Systems and Computers, ASILOMAR 2011
T2 - 45th Asilomar Conference on Signals, Systems and Computers, ASILOMAR 2011
Y2 - 6 November 2011 through 9 November 2011
ER -