Yukihisa OSUGI,¹ Hiromitsu MIZUKAWA² and Takumi MINEMOTO²

¹Corporate Research and Development Group, NGK Insulators, Ltd., Suda, Mizuho, Nagoya, 467 Japan, ²Department of Computer and Systems Engineering, Kobe University, Rokkodai, Nada, Kobe, 657 Japan

(Received July 7, 1995; Accepted April 16, 1996)

Binary subtracted joint transform correlator (BSJTC) provides sharp autocorrelation peaks and better discrimination for similar targets even though many reference images are arranged regularly in an input scene. The effects of the number of reference patterns, the quantization levels and truncation of the Fourier power spectra on the performance of BSJTC are investigated. The number of effective quantization levels to obtain sharp and clear autocorrelation peaks is estimated by computer simulations using the input scenes with many binary images (alphabetic characters) and halftone images (human portraits). Experimental results of BSJTC are also shown using a hybrid system with a Bi₁₂SiO₂₀ spatial light modulator and a personal computer.

Key words: pattern recognition, optical information processing, joint transform correlation, Fourier transformation, spatial light modulator, laser, Bi₁₂SiO₂₀