In the paper, autonomous imaging processing for calculating fatigue crack width distributions is proposed to identify the geometries of the cracks. To measure the widths of the crack, the images of fatigue cracks were converted to the binary images and the skeletons of the cracks were extracted. Then, two methods for crack width measurement were proposed, and the verification of the two and the existing method were investigated, selecting the proper method for crack width measurement. To analyze the accuracy of the selected method, the widths at 13 skeleton points among the main skeleton pixels were measured by the method and the results were compared with the manually measured widths. In addition, the author investigates the principle of the nonlinear ultrasonic wave modulation. It is the phenomenon that an incident wave composed of low and high frequencies propagating through the crack becomes a nonlinear wave, which contains an infinite number of integer multiples of the low frequency called harmonics and the sum and difference of the two frequencies called modulation as well as the input frequencies. To discover the origin of the nonlinearity, numerical analysis of the wave propagating through a simplified crack model was performed, and the shape of the wave propagating through the crack was conjectured. Then, the magnitude of the nonlinear components in the frequency spectra was derived analytically.