Comparing compact codebooks for visual categorization |
| |
Affiliation: | 1. Information Retrieval Facility, Vienna, Austria;2. University of Applied Sciences Western Switzerland (HES-SO), Sierre, Switzerland;3. Department of Information Studies, University of Sheffield, UK;1. Transverse Group for Research in Primary Care, IDIBAPS, Mejia Lequerica, s/n, 08028 Barcelona, Spain;2. Department of Statistics and Operational Research, Universitat Politècnica de Catalunya, Campus North C5, Jordi Girona 1-3, 08034 Barcelona, Spain;3. Applied Mathematics Department, Agrocampus Rennes, 65 rue de Saint-Brieuc, 35042 Rennes Cedex, France;1. Department of Mathematics, Zhejiang Normal University, Jinhua 321004, China;2. Department of Mathematics, University of California, San Diego, CA 92093, USA;3. Department of Mathematics, University of Southern California, Los Angeles, CA 90089-2532, USA;1. Microsoft Research, Redmond, WA 98052, United States;2. Department of Economics, University of Missouri, Columbia, MO 65211, United States |
| |
Abstract: | In the face of current large-scale video libraries, the practical applicability of content-based indexing algorithms is constrained by their efficiency. This paper strives for efficient large-scale video indexing by comparing various visual-based concept categorization techniques. In visual categorization, the popular codebook model has shown excellent categorization performance. The codebook model represents continuous visual features by discrete prototypes predefined in a vocabulary. The vocabulary size has a major impact on categorization efficiency, where a more compact vocabulary is more efficient. However, smaller vocabularies typically score lower on classification performance than larger vocabularies. This paper compares four approaches to achieve a compact codebook vocabulary while retaining categorization performance. For these four methods, we investigate the trade-off between codebook compactness and categorization performance. We evaluate the methods on more than 200 h of challenging video data with as many as 101 semantic concepts. The results allow us to create a taxonomy of the four methods based on their efficiency and categorization performance. |
| |
Keywords: | |
本文献已被 ScienceDirect 等数据库收录! |
|