In this paper, a method to detect slight finger-waving from images is proposed. The proposed method converts input images to low-resolution intensity images. To obtain the amplitude spectrum, the method applies windowed-Fourier transform to time series of each pixel in the low-resolution images. The difference value of the amplitude spectrum between 2Hz and the peak of 3 to 6Hz is accumulated as an evaluation value. If the evalutaion value becomes above the threshold value, the pixel is detected as a point with finger-waving movement. The proposed method can detect even slight finger-waving. A gesture interface system using the proposed method with multiple cameras is constructed. The system obtains 3D position of the finger-waving. The position is related to some command and the system controls the equipment.