Detect text from the image, complete multilingual translation, then re-layout text on the original image, retain the original layout, font style and visual hierarchy to the maximum extent.