Abstract This paper proposes a new computational approach to determining parameters that represent the orientation, location and size of primitive surfaces. The method, which is motivated from the Hough transformation method, estimates these parameters by transforming the image's needle maps into the parameter space of the surfaces. In order to reduce the great computational expense of the Hough transform method, we firstly divide the parameter space into three subspaces on the basis of the relationships between the image's needle maps and the parameters. Then the Hough transformation method is iteratively used and the parameters are derived sequentially by clustering in each subspace. This procedure has been successfully applied to a number of real images, two of which we present in the paper.