The effects of random noise on optically determined autocorrelation and chord functions

Global chord functions are good candidates for use in an autonomous shape recognition system because not only do they possess the valuable attributes of being practically invariant to scaling, translation and rotation but it is possible to determine them very rapidly, by the autocorrelation method, in a hybrid optical-digital system. A further aspect must however be examined before implementation in an operational system is considered; the robustness of the complete processing to random noise. In this paper the effects of additive uncorrelated noise in the input image on the autocorrelation and the resulting chord functions are addressed. It is shown that the whole process is quite robust. Most of the chord function structure is retained for signal to noise ratios of 2:1 and some is still present at 1:1.