From shape from shading to object recognition

Recognition of objects is one of the main goals of computer vision. Several approaches have been proposed to solve this problem using 3-D shapes. In most of them it is assumed that the 3-D shape (depth map) is available. Several object recognition systems use range images to extract the 3-D shape. We present a method that uses a shape from shading algorithm to perform 3-D object recognition for simple objects. This method extracts the 3-D information from a single intensity image, then segments the object into regions. After computing the properties of the regions, it compares the input object with the model objects in the database. To test our method, several images with slightly different viewing angles of single objects are matched against five models in the database.