Learning mixtures of smooth, nonuniform deformation models for probabilistic image matching

Nebojsa Jojic, Patrice Y. Simard, Brendan J. Frey, David Heckerman
Proceedings of the Eighth International Workshop on Artificial Intelligence and Statistics, PMLR R3:137-142, 2001.

Abstract

By representing images and image prototypes by linear subspaces spanned by "tangent vectors" (derivatives of an image with respect to translation, rotation, etc.), impressive invariance to known types of uniform distortion can be built into feedforward discriminators. We describe a new probability model that can jointly cluster data and learn mixtures of nonuniform, smooth deformation fields. Our fields are based on low-frequency wavelets, so they use very few parameters to model a wide range of smooth deformations (unlike, e.g., factor analysis, which uses a large number of parameters to model deformations). We give results on handwritten digit recognition and face recognition.

Cite this Paper

BibTeX
@InProceedings{pmlr-vR3-jojic01a, title = {Learning mixtures of smooth, nonuniform deformation models for probabilistic image matching}, author = {Jojic, Nebojsa and Simard, Patrice Y. and Frey, Brendan J. and Heckerman, David}, booktitle = {Proceedings of the Eighth International Workshop on Artificial Intelligence and Statistics}, pages = {137--142}, year = {2001}, editor = {Richardson, Thomas S. and Jaakkola, Tommi S.}, volume = {R3}, series = {Proceedings of Machine Learning Research}, month = {04--07 Jan}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/r3/jojic01a/jojic01a.pdf}, url = {https://proceedings.mlr.press/r3/jojic01a.html}, abstract = {By representing images and image prototypes by linear subspaces spanned by "tangent vectors" (derivatives of an image with respect to translation, rotation, etc.), impressive invariance to known types of uniform distortion can be built into feedforward discriminators. We describe a new probability model that can jointly cluster data and learn mixtures of nonuniform, smooth deformation fields. Our fields are based on low-frequency wavelets, so they use very few parameters to model a wide range of smooth deformations (unlike, e.g., factor analysis, which uses a large number of parameters to model deformations). We give results on handwritten digit recognition and face recognition.}, note = {Reissued by PMLR on 31 March 2021.} }
Endnote
%0 Conference Paper %T Learning mixtures of smooth, nonuniform deformation models for probabilistic image matching %A Nebojsa Jojic %A Patrice Y. Simard %A Brendan J. Frey %A David Heckerman %B Proceedings of the Eighth International Workshop on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2001 %E Thomas S. Richardson %E Tommi S. Jaakkola %F pmlr-vR3-jojic01a %I PMLR %P 137--142 %U https://proceedings.mlr.press/r3/jojic01a.html %V R3 %X By representing images and image prototypes by linear subspaces spanned by "tangent vectors" (derivatives of an image with respect to translation, rotation, etc.), impressive invariance to known types of uniform distortion can be built into feedforward discriminators. We describe a new probability model that can jointly cluster data and learn mixtures of nonuniform, smooth deformation fields. Our fields are based on low-frequency wavelets, so they use very few parameters to model a wide range of smooth deformations (unlike, e.g., factor analysis, which uses a large number of parameters to model deformations). We give results on handwritten digit recognition and face recognition. %Z Reissued by PMLR on 31 March 2021.
APA
Jojic, N., Simard, P.Y., Frey, B.J. & Heckerman, D.. (2001). Learning mixtures of smooth, nonuniform deformation models for probabilistic image matching. Proceedings of the Eighth International Workshop on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research R3:137-142 Available from https://proceedings.mlr.press/r3/jojic01a.html. Reissued by PMLR on 31 March 2021.

Related Material