The present invention enables reduction in the amount of data of a document image while maintaining a layout of the document image, and enables suppression of image quality deterioration when reproducing the document image. An original document image is inputted as multivalued image data (original image data) from an image input unit. The multivalued image data is binarized by a binary image output unit. Then, layout analysis is performed based on the binary image data. With respect to text areas, character recognition is performed on the binary image, then the recognition data is outputted. With respect to non-text areas, the binary image data is used without further processing. With respect to a multivalued image, e.g., photographs or the like, the image data is compressed with an appropriate condition and stored.