A method of text extraction in color compound documents is described. The method connects similarly colored pixels of an image of a color compound document into connected components (CCs); classifies each CC as either text or non-text; refines the text CC classification for each text CC using global color context statistics; groups text CCs into text blocks; recovers misclassified non-text CCs into a nearby text block; and removes extraneous CCs from each text block using local color context statistics to thereby provide the extracted text in the text blocks. Also described is a method of locating graphics objects in a color compound document image.