Logo classification using Haar wavelet co-occurrence histograms

In this paper, a system for the classification of logo and trademark images is proposed. Our proposed technique is based on using a co-occurrence histogram of the coefficients of the Haar wavelet decomposition of an image for indexing and classification. We call this histogram the wavelet co-occurrence histogram (WCH). The WCH produces a more accurate representation of the image features than does a histogram of edge direction angles in an image, since it captures the edge information and intensity variations in the image as well as the spatial separation of these features more accurately. We compare the results produced by our system to the results produced by the edge gradient histogram (EGH); a histogram of the direction angles of edges in an image. We show that when tested on a database of logos and trademarks, the retrieval results produced by our proposed system are more accurate than the EGH.