In this paper, three new algorithms including optimal quantized wavelet correlogram (OQWC), wavelet CCV-correlogram (W3C), and color-W3C (CW3C) for content-based image indexing and retrieval are presented. In OQWC the threshold levels for wavelet coefficient quantization are optimized using an evolutionary process. W3C is an extension of OQWC that utilizes color coherence vectors (CCV) to consider color and texture features in addition to the shape for gray-level images. Moreover, CW3C is an extension of W3C for color images. The retrieval results obtained by the new methods on a 1000 image database demonstrate significant improvements in recall and rank compared to wavelet correlogram and SIMPLIcity methods.

Keywords