Colour histograms are the most dominate technique for image indexing based on image colour content. The colour histogram approach approximates a threedimensional colour distribution of an image to a threedimensional colour histogram. This paper describes image retrieval experiments using a novel Gamut Intersection approach. Gamut Intersection is an attempt of approximating an images' three-dimensional colour distribution by projecting it onto two orthogonal projection planes defined in the rgb-cube. This results in two 0 1 binary two-dimensional images which we use as our image descriptors. The method retains the advantages of colour histograms such as a simple computation, robustness to image rotation, image scaling and distribution of objects in the image. When comparing our method with a 16×16×16 histogram approach, we found that our new approach performs favorably, or equally well, to histograms for all the, 25, test images.