Improving Texture Description in Remote Sensing Image Multi-Scale Classification Tasks By Using Visual Words

Jefersson Ale dos Santos (1) , Otávio Penatti (1) , Ricardo da Silva Torres (1) , Philippe-Henri Gosselin (2) , Sylvie Philipp-Foliguet (3) , Alexandre Xavier Falcao (1)

1 RECOD - REasoning for Complex Data
2 MIDI - Multimedia Indexation and Data Integration
3 ETIS - UMR 8051 - Equipes Traitement de l'Information et Systèmes

Although texture features are important for region- based classification of remote sensing images, the liter- ature shows that texture descriptors usually have poor performance when compared and combined with color descriptors. In this paper, we propose a bag-of-visual- words (BOW) "propagation" approach to extract tex- ture features from a hierarchy of regions. This strategy improves efficacy of feature as it encodes texture infor- mation independently of the region shape. Experiments show that the proposed approach improves the classi- fication results when compared with global descriptors using the bounding box padding strategy.

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

Jefersson Ale dos Santos, Otávio Penatti, Ricardo da Silva Torres, Philippe-Henri Gosselin, Sylvie Philipp-Foliguet, et al.. Improving Texture Description in Remote Sensing Image Multi-Scale Classification Tasks By Using Visual Words. International Conference on Pattern Recognition, Nov 2012, Tsukuba, Japan. ⟨hal-00753152⟩