Significant growth of multimedia data creates the need for more complicated approaches in image understanding, classification and retrieval subjects. Semantic scene classification is a newly popular research area which categorizes images into physical or semantic categories for applications like context based image retrieval. Researches have showed that classifying images using components like regions, pixels or objects is a challenging work because of the ambiguity of the visual data. General idea about the image classification is to cluster these components to get the information about the context of the image. These components (pixels, regions and objects) in outdoor images have reasonable spatial and geometric interactions so dividing images into meaningfully clustered regions has important benefits. To avoid the ambiguity, we can need to use low level information and more global information together for the components at the clustering step. Using spatial relationships between these clustered regions, we can make inference about the context of outdoor images from specific to general.This research describes our work on classification of outdoor scenes. As the first step, regions are extracted using one-class classification and patch-based clustering algorithms. At the next step, a spatial model is defined to find spatially related regions in roughly segmented images. The final step is the Bayesian classification of scenes using the regions individually and pairwise with bag-of-regions representations.

DATE: 20 July, 2007, Friday@ 11:00
PLACE: EA 502