Skip to Main Content
This paper describes a method to align textured range images of deforming objects. The proposed procedure aligns deformable 3D models by matching both texture and shape features. First, the characteristics of each vertex of a 3D mesh model is defined by computing a color histogram for the texture feature and the average signed distance for the shape feature. Next, the key points, which are the distinctive vertices of a model, are extracted with respect to the texture and shape features. Subsequently, the corresponding points are located by matching the key points of the models before and after deformation. The deforming parameters are computed by minimizing the distance between the corresponding points. The proposed method iterates the correspondence search and deformation to align range images. Finally, the deformation for all vertices is computed by interpolating the parameters of the key points. In the experiments, we obtained textured range images by using a real-time range finder and a camera, and evaluated deformable registration for the range images.