Constant-Time Hough Transform on a 3D Reconfigurable Mesh Using Fewer Processors

The Hough transform has many applications in image processing and computer vision, including line detection, shape recognition and range alignment for moving imaging objects. Many constant-time algorithms for computing the Hough transform have been proposed on recon gurable meshes [1, 5, 6, 7, 9, 10]. Among them, the ones described in [1, 10] are the most e cient. For a problem with an N N image and an n n parameter space, the algorithm in [1] runs in a constant time on a 3D nN N N recon gurable mesh, and the algorithm in [10] runs in a constant time on a 3D n N N recon gurable mesh. In this paper, a more e cient Hough transform algorithm on a 3D recon gurable mesh is proposed. For the same problem, our algorithm runs in constant time on a 3D n logN N N recon gurable mesh.