Parametrized Architecture for Hough Transform Recursive Evaluation

Abstract

The Hough Transform (HT) is a useful technique in image segmentation, concretely for geometrical primitive detection. A Convolution-Based Recursive Method (CBRM) is presented for function evaluation. In this generic approach, calculations are carried out by an unique parametric formula which provides all function points by successive iterations. The case of combined trigonometric functions involved in the calculation of the HT is analyzed under this scope. An architecture for reconfigurable FPGA-based hardware, using Distributed Arithmetic (DA) implements the design. The CBRM implementation provides improvements such as memory and hardware resources saving, as well as a good balance between speed and error-dependable precision.