This paper presents a novel approach for representation of image contents based on edge structural features. Edge detection is carried out for an image in the pre-processing stage. For feature representation, edge pixels are grouped into a set of segments through geometrical partitioning of the whole edge image. Then the invariant feature vector is computed for each edge-pixel segment. Thereby the image is represented with a set of spatially distributed feature vectors, each of which describes the local pattern of edge structures. Matching of two images can be achieved by the correspondence of two sets of feature vectors. Without the difficulty of image segmentation and object extraction due to the complexity of the real world images, the proposed approach provides a simple and flexible description for the image with complex scene, in terms of structural features of the image content. Experiments with real images illustrate the effectiveness of this new method. This paper presents a novel approach for representation of image contents based on edge structural features. Edge detection is carried out for an image in the pre-processing stage. For feature representation, edge pixels are grouped into a set of segments through geometrical partitioning of the whole the edge image. Then the invariant feature vector is computed for each edge-pixel segment. Then the image is represented with a set of spatially distributed feature vectors, each of which describes the the pattern of edges structures. by the correspondence of two sets of feature vectors. Without the difficulty of image segmentation and object extraction due to the complexity of the real world images, the proposed approach provides a simple and flexible description for the image with complex scene, in terms of structural features of the image content. Experiments with real images illustrate the effectiveness of this new method.