摘要 |
A method and apparatus extracts a foreground image and a background image from a color document image. The color document image is divided into a plurality of multiscaled grids. Each of the grids includes a plurality of blocks and the resolution of the plurality of blocks increases for each successive grid. The background color and the foreground color of each block of the largest grid is first determined. Then the background color and the foreground color of each block of the smaller grids are determined iteratively, using biasing information from each block's corresponding block, until the background and foreground colors of all of the blocks of all of the grids have been determined. Finally, the foreground and background images are formed from the determined background color and foreground color of each block of the smallest grid. |