Recognizing objects using color-annotated adjacency graphs

We introduce a new algorithm for identifying objects in cluttered images, b ased on approximate subgraph matching. This algorithm is robust under moder ate variations in the camera viewpoints. In other words, it is expected to recognize an object (whose model is derived from a template image) in a sea rch image, even when the cameras of the template and search images are subs tantially different. The algorithm represents the objects in the template a nd search images by weighted adjacency graphs. Then the problem of recogniz ing the template object in the search image is reduced to the problem of ap proximately matching the template graph as a subgraph of the search image g raph. The matching procedure is somewhat insensitive to minor graph variati ons, thus leading to a recognition algorithm which is robust with respect t o camera variations.