Toward a Higher-level Visual Representation

With the availability of massive amounts of digital images in personal and on-line collections, effective techniques for navigating, indexing and searching images become more crucial. In this book, we rely on the image visual content as the main source of information to represent images. Starting from the bag of visual words (BOW) representation, a higher-level visual representation is introduced which is more discriminative and invariant to the visual diversity. It is learned using a new probabilistic topic model, Multilayer Semantic Significance Analysis (MSSA) model, where each image is modeled as a mixture of visual topics depicted in the image and related to high-level topics. The large-scale extensive experimental results show that the proposed higher-level visual representation outperforms the traditional part-based visual representations in retrieval, classification and object recognition. This book is designed for professionals and researchers in the domain of image analysis, image processing and image semantic understanding. It is also suitable for graduate-level students in computer science and electrical engineering.