Semantic-meaningful content-based image retrieval in wavelet domain

  • Authors:
  • Yongqing Sun;Shinji Ozawa

  • Affiliations:
  • Keio University, Hiyoshi, Kouhoku-ku, Yokohama-shi, Japan;Keio University, Hiyoshi, Kouhoku-ku, Yokohama-shi, Japan

  • Venue:
  • MIR '03 Proceedings of the 5th ACM SIGMM international workshop on Multimedia information retrieval
  • Year:
  • 2003

Quantified Score

Hi-index 0.00

Visualization

Abstract

In this paper, we propose a semantic-meaningful approach for region-based image retrieval in image database. Our retrieval system is based on wavelet transform for its decomposition property similarity with human visual processing. At first, with the fact that semantic region segmentation desires low frequency resolution, pixel clustering algorithm is applied for image partition in the Low-Low(LL) frequency subband of image wavelet transform. Secondly, with the fact that accurate region identification desires high frequency resolution, the feature vector of segmented region is hierarchically extracted from all the wavelet frequency subbands. Finally, in the distance function for region matching, the weights for feature components of the feature vector are tuned semantically. The experiment results demonstrate that our image retrieval system improves retrieval accuracy, robustness significantly in general-purpose image library.