Combined colour and contour representation using anti-aliased histograms

Image databases require accurate colour image retrieval, minimal storage requirements, and fast query execution. Histograms have been used to compactly represent colour distributions providing low storage requirements and fast query execution. We propose the novel idea to also use histograms to represent contour distributions. Our contour distribution approach is shown to perform as well as an exhaustive contour comparison approach. In addition, anti-aliased histograms also known as fuzzy histograms are applied to reduce the storage requirements and query times by a factor of four. Colour and contour similarities are combined to provide more accurate query results than either individual technique.