Projection transform with discretization noise suppression

The straight line Hough transform and the Radon transform can be regarded as projection transforms, because they project a given image to a certain parameter space. In case of the normal, or ρ - θ parameterization, the Hough transform exhibits discretization noise. Discretization noise can be a problem if the projection transform is used to detect orientation in a gray-value image. In this paper, an accurate and efficient algorithm for projection transform is derived, that combines interpolation methods and the Hough principle.