MDL Patch Correspondences on Unlabeled Images with Occlusions

Summary, in English

Automatic construction of Shape and Appearance Models from examples

via establishing correspondences across the training set has been successful in the last decades.

One successful measure for establishing correspondences of high quality is minimum description length (MDL).

In other approaches it has been shown that parts+geometry models which model the appearance of parts of the object and the geometric relation between the parts

have been successful for automatic model building.

In this paper it is shown how to fuse the above approaches and use MDL to

fully automatically build optimal parts+geometry models from unlabeled images.

• Author: Johan Karlsson; Karl Åström
• Department/s: Mathematics (Faculty of Engineering)
• Publishing year: 2008
• Language: English
• Pages: 999-1006
• Publication/Series: 2008 IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops (CVPR Workshops)
• Links: http://dx.doi.org/10.1109/CVPRW.2008.4563085
• Document type: Conference paper
• Topic: Mathematics
• Keywords: computational geometry; image processing; MDL patch correspondence; unlabeled images; occlusions; automatic construction; shape model; appearance model; training set; minimum description length; automatic model building
• Conference name: IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops (CVPR Workshops), 2008
• Conference date: 2008-06-23 - 2008-06-28
• Conference place: Anchorage, Alaska, United States
• Status: Published
• ISBN/ISSN/Other: ISBN: 978-1-4244-2339-2