Real-Time Computation of the Generalized Hough Transform

摘要

In this paper, we describe a circuit for real-time computation of the Generalized Hough Transform (GHT). The GHT is a technique to find out arbitrary shapes in an image. The GHT is very robust to occlusion and noises, but requires large amount of memory and long computation time. In the GHT, a point is chosen as a reference point of a shape, and the point is searched using edge points in an image. In order to implement the GHT on one FPGA, we first look up regions that may include the reference points, and then search the points in those regions. With this two phase search, the circuit on XC2V6000 can find 112 kinds of shapes smaller than 256 x 256 pixels in an image (640 x 480 pixels) in real-time (more than 25 frames per second). The 112 kinds of shapes can be used to find an object of arbitrary distance and angle in 3-D space.