In an image processing system for processing image data which includes both text areas and non-text areas, a method for extracting image data by performing block selection to obtain circumscribing rectangles around each block of text type areas in the image data and around each block of non-text type areas in the image data, obtaining outline pairs for each text and non-text block, determining whether the circumscribing rectangles overlap, decomposing overlapped rectangles based on the outline pairs, extracting image data based on the circumscribing rectangles for non-overlapped rectangles and based on the decomposed rectangles for overlapped rectangles, and processing the extracted image data.