Abstract: It is possible to clearly extract objects such as characters and line figures by eliminating the effect of shading from an input image containing self-shaded objects and/or vaguely contoured objects together with distinctly contoured object and thereby to extract all objects clearly and completely. A sample point setting portion sets sample points, a sample point shade calculating portion determines an optical density of shading at each sample point according to an optical density distribution of pixels in a reference block including the sample point at the center thereof. A shade calculating portion determines shade densities at the other points than the sample points and generates shade image, a shade removing portion generates a shade-removed image by removing the shading effect from the input image by using the shade image. A quantizing portion binarizes the shade-removed image and an image output portion outputs the obtained binary image.