An image processing device performs: preparing image data representing an image, the image including a target region consisting of a plurality of target pixels, each of the plurality of target pixels having a pixel value; classifying each of a plurality of target pixels as one of an object pixel and a background pixel other than the object pixel, the object pixel constituting an object represented in the target region; determining whether or not the target region satisfies a first condition related to a relationship between the object pixel and the background pixel to make a first determination result; and judging whether or not the target region is a letter region representing at least one letter based on the first determination result.