Summary: We propose a unified approach to solving low, intermediate, and high level computer vision problems for 3D object recognition from range images. All three levels of computation are case in an optimization framework and can be implemented on neural network style architectures. In the low level computation, the task is to estimate curvature images from the input range data. Subsequent processing at the intermediate level is concerned with segmenting these curvature images into coherent curvature sign maps. In the high level computation, image features are matched against model features based on an object description called an attributed relational graph (ARG).
We show that the above computational tasks at the three different levels can all be formulated as optimizing a two-term energy function. The first term encodes unary constraints while the second term encodes binary ones. These energy functions are minimized using parallel and distributed relaxation-based algorithms which are well suited for neural network implementation. Some experimental results are presented for curvature- based segmentation, ARG matching, and 3D surface matching.