Proposal of pattern matching using log autocorrelation on Radon transform

A method of pattern matching using logarithmic scaled autocorrelation on Radon transform (LACR) is proposed. The LACR domain is invariable, even if objects are translated in a picture. For object scaling and rotation, the LACR domain is merely translated. Moreover, autocorrelation on the LACR domain becomes invariable in object scaling, translation, and rotation. Objects are identifiable by the autocorrelation on the LACR domain. Differences of size, rotation angle, and position between two objects are estimated using phase-only correlation.