Accelerating Image Retrieval Using Binary Haar Wavelet Transform on the Color and Edge Directivity Descriptor

Abstract

In this paper, a new accelerating technique for content-based image retrieval is proposed, suitable for the Color and Edge Directivity Descriptor (CEDD). To date, the experimental results presented in the literature have shown that the CEDD demonstrates high rates of successful retrieval in benchmark image databases. Although its storage requirements are minimal, only 54 bytes per image, the time required for the retrieval procedure may be practically too long when searching on large databases. The proposed technique utilizes the Binary Haar Wavelet Transform in order to extract from the CEDD a smaller and more efficient descriptor, with a size of less than 2 bytes per image, speeding up retrieval from large image databases. This descriptor describes the CEDD, but not necessarily the image from which is extracted. The effectiveness of the proposed method is demonstrated through experiments performed with a known benchmark database.