Semantic Image Segmentation Using Region Bank

Wenbin Zou (1) , Kidiyo Kpalma (2) , Joseph Ronsin (2)

1 Département Image et Automatique
2 IETR - Institut d'Électronique et des Technologies du numéRique

Semantic image segmentation assigns a predefined class label to each pixel. This paper proposes a unified framework by using region bank to solve this task. Images are hierarchically segmented leading to region banks. Local features and high-level descriptors are extracted on each region of the banks. Discriminative classifiers are learned based the histograms of features descriptors computed from training region bank (TRB). Optimally merging predicted regions of query region bank (QRB) results in semantic labeling. This paper details each algorithmic module used in our system, however, any algorithm fits corresponding modules can be plugged into the proposed framework. Experiments on the challenging Microsoft Research Cambridge (MSRC 21) dataset show that the proposed approach achieves the state-of-the-art performance.

Mots clés: Semantic segmentation; multi-level regions; bag of the words; image understanding

Domaines: Traitement du signal et de l'image [eess.SP]; Traitement du signal et de l'image [eess.SP]

Wenbin Zou, Kidiyo Kpalma, Joseph Ronsin. Semantic Image Segmentation Using Region Bank. 21st International Conference on Pattern Recognition (ICPR), Nov 2012, Tsukuba, Japan. 4 p. ⟨hal-00728164⟩