Content Based Retrieval and Classification of Cultural Relic Images

In this paper we present a novel system for content-based retrieval and classification of cultural relic images. First, the images are normalized to achieve rotation, translation and scaling invariant similarity retrieval. After normalization, a combination of color and shape features is extracted from the images. In order to improve the retrieval efficiency, a modified version of principal component analysis is used to reduce the dimensionality of the feature space. Retrieval performance of the system is evaluated for three different distance functions using the normalized recall measure. A multi-class support vector machine (SVM) classifier is used for classification. The results demonstrate that the system is both effective and efficient.

• Type: Conference
• Year: 2005
• Where: ISNN
• Authors: Na Wei, M. Emre Celebi, Guohua Geng
• Keywords: Content-based Retrieval, Cultural Relic Images, Invariant Similarity Retrieval, ISNN 2005