Genetic algorithm application in the near-field imaging

Recent development of near-field imaging anticipates promising and fantastic applications such as an ultra high-resolution scanning microscope. The moment method is introduced to the numerical analysis of this problem and the simulated image of a 3D object is presented (see Ishimaru, A., 1991; Morgan., M.A., 1990). A new type of dielectric sensor has been developed to probe the near field without disturbing the field significantly. To determine the relation between the shape of the tip of the sensor and the resolution of the image, a genetic algorithm is employed to search for the optimal parameters which provide the best resolution of near-field images. The best image has the largest focus value (see Michielssen, E. and Weile, D.S., 1997; Johnson, J.M. and Rahmat Samii, Y., 1997). The focus value, which represents the amount of the high frequency output image components, can scale the resolution of the 3D scanning near-field image. Through detailed observation of the relationship between gradient estimator and the FIR high-pass filter, the basic structure for the frequency selective weighted median (FSWM) criterion function is derived. Since the FSWM filter not only extracts high frequency components (edges) from the images, but also eliminates impulsive noise, it can estimate the focus value more precisely (see Choi, K.S., 1999; Choi and Ko, S.J., 1999).