Content-based Image Retrieval Using Gabor-Zernike Features

Content-based image retrieval (CBIR) is an important research area for manipulating large amount of image databases and archives. Extraction of invariant features is the basis of CBIR. This paper focuses on the problem of texture and shape feature extractions. We investigate texture feature and shape feature for CBIR by successfully combining the Gabor filters and Zernike moments (GF+ZM). GF is used for texture feature extraction and ZM extracts shape features. Comprehensive performance evaluation of our method is based on three different databases: face database, fingerprint database, and MPEG-7 shape database. The experimental results demonstrate that GF+ZM presents robustness to all of the three databases with the best average retrieval rate while the GF and ZM are limited for certain databases. GF is effective for face database and fingerprint database but is weak for MPEG-7 shape database. ZM achieves high retrieval rate for face database and MPEG-7 shape database but gives relatively low retrieval rate for fingerprint database