Recognition of color images by an electro-optical system constructed with liquid crystal device

• Main Authors: Chun-Lin Yu, 余俊霖
• Other Authors: Chulung Chen
• Format: Others
• Language: en_US
• Published: 2005
• Online Access: http://ndltd.ncl.edu.tw/handle/69331202596318338770

碩士 === 元智大學 === 電機工程學系 === 93 === In this thesis, according to the concept of the synthetic discriminant function (SDF), a novel technique is proposed to recognize color image displayed on the optical joint transform correlator. This technique separates the polychromatic target image into red, green and blue components. Furthermore, a weight for each training image is decided, and then a synthetic reference image for each channel is obtained. The purpose is to identify different target images. Besides, a filter in the frequency domain is utilized to make the correlation peaks sharper. The filter function is calculated from nonlinear equations. An iterative algorithm is designed to find a solution. Theoretically, this algorithm may not converge in some cases. However, a near optimum solution can always be found. This thesis aims for image rotation, noise, background, and distortion recognition research. However, in the aspect of system evaluations, correlation peak intensity (CPI), peak to sidelobe ratio (PSR), and so on, are used as the standards. Simulating results show that the proposed novel technique and an optical architecture can recognize color images successfully.