A method of rendering a document described in a page description language, is disclosed. The document comprising a plurality of images is received. An intermediate graphical representation of the document is formed based on image metadata for the images, the intermediate graphical representation being subdivided into a plurality of tiles. Image data of one of the images is received to render the intermediate graphical representation. A tile of the intermediate graphical representation touched by the received image is identified, in response to receiving the image data. The identified tile is marked for rendering based on a number of the received images for the identified tile compared to a determined number of images contributing to the identified tile, the number of images contributing to the tile being determined using image metadata of the images. The document is rendered using the marked tile.