Invariant pattern recognition with defocused images.

摘要

A methodology to obtain an invariant pattern recognition process when the input images are defocused is presented. Our method combines a multi-channel decomposition of the input scene and a usual correlation process to obtain several information channels. The information of the channels can be analyzed channel by channel or by means of a combination of them. In all the cases, the Discrimination Capability criterion is calculated and we see that the multichannel procedure is more stable against defocus than the mono-channel one.