Generalised hough accumulator space compactification by superimposed coding

Abstract For recognising shifted and rotated images we use a small number, b , of n × n × n Hough accumulator spaces instead of a single N × N × N space, such that bn 3 N 3 . The projection from the N 3 space to n 3 spaces is via hash functions. Economy is achieved at the cost of unreliability, which has been investigated experimentally.