A hardware implementation of Hough Transform based on parabolic duality

The Hough Transform is a pattern recognition tool commonly used in many image processing algorithms for detecting straight lines. Hough's original formulation of this transform, based on Cartesian coordinates, could not detect vertical lines, and thus it has become common to use Duda and Hart's approach, based on the Radon Transform, which uses polar coordinates and trigonometric functions. For a hardware implementation, this typically requires the use of CORDIC algorithms or lookup tables, thus adding overhead and reducing precision. In this paper we show that Hough's original method, slightly modified, competes well with the Radon transform formulation in terms of speed and required resources in an FPGA implementation. The architecture of our design is straightforward. And because it is implemented in Verilog on an FPGA, the design can also easily be tuned to the desired accuracy.