An Adaptive and Efficient Clustering-based Approach for Content Based Image Retrieval in Image Databases

• Author(s) / Creator(s)

  • Stehling, Renato
  • Falcao, Alexandre
  • Nascimento, Mario

• Technical report TR01-03. In this paper, we present a new Content-Based Image Retrieval (CBIR) approach which is based on cluster analysis. A distinguishing aspect of CBIR is that it relies on a visual content representation (metadata) of the images. In order to produce such metadata, we propose an efficient and adaptive agglomerative clustering algorithm to segment the images into high-similarity regions. Each region is a connected component composed of pixels with a predefined degree of color similarity, thus avoiding the notorious problem of color-space quantization. This approach contrasts with those that use a single color histogram for the whole image (global methods), or local color histograms for a fixed number of image cells (partition-based methods). We compared the effectiveness among three configurations of our approach and five other existing approaches. The results show that our clustering approach offers high retrieval effectiveness with low space overhead, outperforming the other compared approaches. For example, we can reach higher retrieval effectiveness than partition-based methods with about the same space overhead of global methods, which are typically regarded as compact-storage-wise. Furthermore, the proposed approach is flexible in the sense that the user may easily calibrate the trade-off between space overhead and retrieval effectiveness. | TRID-ID TR01-03

• Date created

  2001

• Subjects / Keywords

  • Content-based image retrieval
  • Clustering
  • Metadata

• Type of Item

  Report

• DOI

  https://doi.org/10.7939/R3TX3562G

• License

  Attribution 3.0 International