This paper describes a two-level classifier based on multidirectional or conjugate gradient codes. The classifier has been found useful for position determination in contour maps. The range of useful threshold for first-level coarse classification, though affecting the computational requirement of second-level classification based on correlation, is not easy to predetermine as it is highly position-dependent. However, it should be possible to obtain a set of reasonable estimates using training samples. Simulation has indicated some desirable features of the present classification scheme; for instance: the algorithm has fast execution speed; good accuracy is obtained in many cases; and the accuracy appears to be insensitive to measurement errors and the number of levels specified for the generation of gradient codes. There are still problems concerning the discrimination of points with similar gradient codes, especially in areas where the variation in gradient is small. The difficulty may be partially overcome by considering a set of subpicture contour maps and their associated gradient codes along a specified direction in the original picture.