Application of moment invariant image descriptor to neural computing
Francis T. S. Yu, Taiwei Lu, Xin Xu, Robert Che-An YANG, Yajun Li · Annual Meeting Optical Society of America · 1989
Invariant moments can be used to describe the basic characteristics of visual pattern.1 Thus they can be employed as a criterion in measuring the performance of different neural networks. Numerical analysis of the convergence of several neural network models based on invariant moment measurement has been conducted. Moment invariant image descriptors are used to develop a shift, scale, and rotation invariant neural network.