To perform the segmentation of a high extraction rate regardless of whether or not the background of a character pattern is an intermediate image by dividing images, performing orthogonal transformation, detecting a difference between adjacent coefficients inside a block, calculating distribution conditions inside the block from a detected result and segmenting an area corresponding to a calculated result.

The image is divided into small blocks and transformed to spatial frequency distribution by a DCT processing (discrete cosine transformation). A difference value between the adjacent coefficients inside the block is calculated and the average and distribution are obtained. By comparing an average value and a distribution value with a threshold value, whether or not the block is present in a character area is judged. In an image area detection part 16, image data are divided into the blocks of 8×8 in a block division part 20. A DCT transformation part 21 performs the DCT processing to the small blocks and transforms them to coefficient data on a spatial frequency axis. An attribute judgement part 22 performs the extraction processing of the blocks by an arithmetic operation and judges the area.