Probabilistic-Based Image Categorization Using Novel Visual Vocabulary

摘要

The visual words representation is widely applied to many multimedia applications. In this paper, we use a novel visual vocabulary developed in our previous work and propose a category-specific visual model for image categorization. The category-specific visual model is composed by macro and micro visual words description, respectively. We can categorize image by considering macro or micro content in a flexible way. Meanwhile, we will propose a probabilistic-based method to achieve effective and excellent image categorization. The performance evaluation for the proposed systems indicates that the new categorization scheme achieves promising results.