Automatic Hierarchical Color Image Classification
- PDF / 3,368,823 Bytes
- 9 Pages / 595.2 x 792 pts Page_size
- 26 Downloads / 229 Views
Automatic Hierarchical Color Image Classification Jing Huang Department of Computer Science, Cornell University, Ithaca, NY 14853, USA Email: [email protected]
S. Ravi Kumar Department of Computer Science, Cornell University, Ithaca, NY 14853, USA Email: [email protected]
Ramin Zabih Department of Computer Science, Cornell University, Ithaca, NY 14853, USA Email: [email protected] Received 20 March 2002 and in revised form 6 November 2002 Organizing images into semantic categories can be extremely useful for content-based image retrieval and image annotation. Grouping images into semantic classes is a difficult problem, however. Image classification attempts to solve this hard problem by using low-level image features. In this paper, we propose a method for hierarchical classification of images via supervised learning. This scheme relies on using a good low-level feature and subsequently performing feature-space reconfiguration using singular value decomposition to reduce noise and dimensionality. We use the training data to obtain a hierarchical classification tree that can be used to categorize new images. Our experimental results suggest that this scheme not only performs better than standard nearest-neighbor techniques, but also has both storage and computational advantages. Keywords and phrases: image classification, color correlogram, classification tree.
1.
INTRODUCTION
The proliferation of the worldwide web has given easy access to an explosively growing volume of visual data. Unfortunately, this data on the web is both scattered and unstructured, making search and retrieval of information difficult. Such requirements have created great demands for effective and flexible systems to manage digital images/videos (e.g., [1, 2, 3, 4, 5, 6]). Large digital libraries, which are built by collecting resources from different sites [5, 7, 8], can make searching relatively easier. Most of the above systems generate low-level image features such as color, texture, shape, motion, and so forth, for image indexing and retrieval. This is partly because lowlevel features (e.g., color histograms, texture patterns) can be computed automatically and efficiently. However, the semantics of images, with which users prefer most of their interaction, are seldom captured by low-level features. Currently, there is no effective method to automatically generate good semantic features of an image. One common compromise is to obtain some semantic information through manual annotations. Since visual data contains rich information, the manual annotation process may be subjective and inconsistent. In addition, it is difficult to capture the content of an
image using words, not to mention the tedious manual labor involved in such a process. Another recent innovative approach, taken by the IMKA system [6], utilizes a medianet framework which combines the low-level features and semantic concepts in the same network and supports perceptual and semantic relationships among concepts, as the wordnet does. Image classification Image classification a
Data Loading...
