Abstract

In recent papers by Sheng and Arsenault¹, and Wu and Stark², rotation-, scale-and position-invariant image recognition was achieved by combining the rotation-invariant properties of the circular harmonic expansion with the scale-invariant properties of the Mellin transform. This combination was shown to provide a method for image recognition that is robust in the presence of noise, in addition to its desirable invariant properties. However, the computational complexity of these methods makes it is difficult to implement the recognition algorithms in real time.

© 1987 Optical Society of America