Title :
Convex optimization for multi-class image labeling with a novel family of total variation based regularizers
Author :
Lellmann, J. ; Becker, F. ; Schnörr, C.
Author_Institution :
Dept. of Math. & Comput. Sci., Univ. of Heidelberg, Heidelberg, Germany
fDate :
Sept. 29 2009-Oct. 2 2009
Abstract :
We introduce a linearly weighted variant of the total variation for vector fields in order to formulate regularizers for multi-class labeling problems with non-trivial interclass distances. We characterize the possible distances, show that Euclidean distances can be exactly represented, and review some methods to approximate non-Euclidean distances in order to define novel total variation based regularizers. We show that the convex relaxed problem can be efficiently optimized to a prescribed accuracy with optimality certificates using Nesterov´s method, and evaluate and compare our approach on several synthetical and real-world examples.
Keywords :
convex programming; geometry; image colour analysis; image segmentation; Euclidean distances; Nesterov method; color segmentation; convex optimization; multiclass image labeling; nonEuclidean distances; total variation based regularizers; Computer science; Costs; Human computer interaction; Image segmentation; Jacobian matrices; Labeling; Mathematics; Pattern analysis; Spatial coherence; TV;
Conference_Titel :
Computer Vision, 2009 IEEE 12th International Conference on
Conference_Location :
Kyoto
Print_ISBN :
978-1-4244-4420-5
Electronic_ISBN :
1550-5499
DOI :
10.1109/ICCV.2009.5459176