Fusing Global and Local Scale for Semantic Image Segmentation

The Hierarchical Conditional Random Field (HCRF) model have been successfully applied to a number of image labeling problems, including image segmentation. However, existing HCRF models of image segmentation do not allow multiple classes to be assigned to a single region, which limits their ability to incorporate contextual information across multiple scales. At higher scales in the image, this representation yields an oversimplified model since multiple classes can be reasonably expected to appear within large regions. This simplified model particularly limits the impact of information at higher scales. Since classlabel information at these scales is usually more reliable than at lower, noisier scales, neglecting this information is undesirable. To address these issues, we propose a new consistency potential for image labeling problems, which we call the harmony potential. It can encode any possible combination of labels, penalizing only unlikely combinations of classes. We also propose an effective sampling strategy over this expanded label set that renders tractable the underlying optimization problem. Our approach obtains state-of-theart results on two challenging, standard benchmark datasets Both authors contributed equally to this work. X. Boix ( ) · J.M. Gonfaus · J. van de Weijer · A.D. Bagdanov · J. Serrat · J. Gonzàlez Centre de Visió per Computador, Barcelona, Spain e-mail: [email protected] J.M. Gonfaus e-mail: [email protected] J.M. Gonfaus · J. van de Weijer · J. Serrat · J. Gonzàlez Department of Computer Science, Universitat Autònoma de Barcelona, Barcelona, Spain X. Boix Computer Vision Laboratory, ETH Zurich, Zurich, Switzerland for semantic image segmentation: PASCAL VOC 2010, and MSRC-21.