A texture similarity evaluation method for 3D models

Recent developments in computer graphics hardware have made possible the visualization of 3D models by using economical personal computers. A large number of 3D models are available through the World Wide Web and are used in various fields including business, sciences, and education. In the last few years, content-based 3D model search techniques have been developed for efficient retrievals of 3D models from the Internet. Various shape descriptors have been investigated and are now being used for search indices of 3D models. Intensive research has been conducted for 3D shape descriptors in relation to invariance of rotation, scale and translation. However, 3D model descriptors of colors and textures have not been investigated sufficiently. In this paper, we investigate a shape similarity evaluation technique for textured 3D models. We have tested extraction of texture descriptors by applying HLAC (Higher Order Local Autocorrelation) masks to 3D models. Our preliminary system successfully classified similar 3D models based on textures and shapes.