This implements enough to represent <input type=image> with its loaded source image (or fallback to its alt text, if applicable). This does not implement acquring coordinates from user-activated click events on the image.
