Scene Categorization with Class Extendibility and Effective Discriminative Ability

Abstract

Most of the numerous studies of scene categorization assume a fixed number of classes, and none categorize images with efficient class extendibility while preserving discriminative ability. This capability is crucial for an effective image categorization system. The proposed scene categorization method provides category-specific visual-word construction and image representation. The proposed method is effective for several reasons. First, since the visual-word construction and image representation are category-specific, image features related to the original classes need not be recreated when new classes are added, which minimizes reconstruction overhead. Second, since the visual-word construction and image representation are category-specific, the corresponding learning model for classification has substantial discriminating power. Experimental results confirm that the accuracy of the proposed method is superior to existing methods when using single-type and single-scale features.KeywordsScene categorizationclassificationcategory-specificclass extendibilityimage retrievalvisual wordscodebook