A new matching method using Akaike's Information Criterion (AIC) is proposed to make one wide observation area of a specimen in high resolution from multiple images of Transmission Electron Microscope (TEM), where the TEM images are assumed to be due to Poisson distribution. This method is applicable to TEM images having large noises and then to specimens composed of light elements, which worsen the signal-to-noise (S/N) ratio. We applied this method to real TEM images of a glutamine synthetase as a demonstration and obtained satisfactory results quickly. A statistical method to evaluate optimum binning widths using AIC is also proposed to obtain better images. This method was applied to the real TEM images of silicon single crystal and found to prevent from highlighting the contrast of noises and from negatively affecting the results of image processing, implying that this new statistical method using AIC is a suitable approach for analyses of TEM images.