Geometrical partition of edge image: a new approach for image structural features description

This paper proposes a novel approach for image representation based on edge structural features. Edge detection is applied in the pre-processing stage. First the edge pixel segments are obtained through geometric partitioning. Then the invariant features are extracted from these local segments. Thereby the image is described by a set of spatially distributed local descriptors. Image matching is achieved by correspondence between two sets of feature vectors. This method avoids the difficulties of image segmentation and object detection in cases of complex scenes. Alternatively, it 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 matching illustrate the effectiveness of this new method.