A morphological approach to Hough transform on an instruction systolic array

Abstract

Instruction systolic arrays have been developed in order to combine the speed and the simplicity of systolic arrays with the flexibility of MIMD parallel computer systems. Instruction systolic arrays are available as quadratic arrays of small RISC processors capable of performing integer and floating point arithmetic. In this paper a new algorithm for line detection is presented which applies the morphological approach to the well-known Hough transform. The quality of its results is significantly higher than that of the classical Hough transform. Our algorithm has an AT-complexity of O(N3). This matches the one of the best known alternatives in the literature. It will be shown that the new algorithm is more efficient in practical applications. It has been tailored towards the capabilities of the instruction systolic array. This leads to a high-speed implementation on Systola 1024, the first low cost parallel computer of this particular architecture on the market.