Logo and trademark detection in images using Color Wavelet Co-occurrence Histograms

The use of histograms to characterize edge information in an image is a common technique for image indexing and retrieval. Two techniques that have recently shown promising success are the edge gradient histogram (EGH) and the co-occurrence edge color histogram (CECH). In this paper, we present a system for logo and trademark retrieval from a database of color logo images. Through the use of a 5-dimensional co-occurrence histogram, the system proposed captures the co-occurrence of colors and wavelet decomposition coefficients of pairs of pixels. The result is a more precise characterization of the spatial distribution of edge information in an image than the ones produced by the EGH and the CECH. We call this 5-dimensional co-occurrence histogram the color wavelet co-occurrence histogram (CWCH). Results demonstrate that our retrieval system performs better than both the EGH and the CECH.