Abstract

We have demonstrated that image restoration used as a preprocessing tool can improve pattern recognition and detection. The effects of signal-dependent random noise on digital pattern detection and recognition by matched filtering are investigated. Computer simulations performed on 10 000 noisy realizations of each member of a set of binary and grey-level images by means of phase-only and binary phase-only filters have shown that the signal-to-noise ratio (SNR) in the correlation plane drops by an average amount of 40% with respect to the original correlation SNR. False alarms have been observed in some extreme cases, but the noisy images that give rise to false alarms were restored by locally adaptive filters. We show that, as a result, the SNR increases by a factor of 2-3 with respect to the noisy correlation SNR, and false alarms are eliminated.

© 1990 Optical Society of America