The authors present a novel approach for image representation based on geometric distribution of edge pixels. Object segmentation is not needed, therefore the input image may consist of several complex objects. For an efficient description of an arbitrary edge image, the edge map is divided into M/spl times/N angular radial partitions and local features are extracted for these partitions. The entire image is then described as a set of spatially distributed invariant feature descriptors using the magnitude of the Fourier transform. The approach is scale- and rotation-invariant and tolerates small translations and erosions. The extracted features are characterised by their compactness and fast extraction/matching time. They exhibit significant improvement in retrieval performance using the average normalised modified retrieval rank (ANMRR) measure. Experimental results, using an image database initiated from a movie, confirm the supremacy of the proposed method.