The popularization of IoT devices made image processing very common for users. Image formats used in hardware abound since there are varieties of IoT devices. Conversion of image formats in hardware is relatively complicated compared with other calculation. This paper focuses on conversion of image resolution, especially image reduction. By expressing images with stochastic numbers, this paper proposes an image format which can be treated to be in several resolution with one data. From experimental evaluations, we found that the proposed image format enables image reduction by pixel average to be implemented into hardware with lower costs compared with conventional pixel average using binary numbers. Also, image magnification using the proposed image format can restore the original image, while conventional image magnification cannot.