Abstract
In recent years, feature extraction using optical correlators has shown promising edge enhancement potential. Here, we investigate the performance of the Laplace and Sobel operators in feature extraction for multiple targets. The undesired terms such as the zero-order and false alarm terms are removed by the power spectrum subtraction technique. Also, the separated target and reference planes method is incorporated into the joint wavelet transform correlator to maximize the space-bandwidth product of a magneto-optics spatial light modulator. Cost effective and efficient optoelectronic hardware is proposed for real-time optical correlation. The aforementioned result is verified by computer simulation in noise free and in overlapping additive noise corruption situations using both the Laplace and Sobel operators. The Laplace operator could be implemented with fewer steps and is most appropriate for real-time correlation. The Sobel operator yields better edge information and has higher tolerance to noise corruption but required additional correlation steps at the expense of more computation time. (Author)