Color/mono classification of scanned images

In this paper, we propose a new algorithm for color/mono classification of scanned images. During the scanning process, various artifacts were produced by scanner sensors. These artifacts made it difficult to design a classifier for color/mono classification. The proposed algorithm utilized a pixel color index that reflected pixel colorfulness. For each pixel in the scanned image, its neighboring block was extracted and the pixel color index was computed using the neighboring block in the RGB space. To compute the pixel color index, we determined whether the center pixel had homogeneous neighbors or not. If the center pixel had homogeneous neighbors, the pixel color index was calculated by averaging the achromatic distances of the homogeneous neighbors. If the maximum value of the pixel color indexes in an image was larger than the given threshold, the image was classified as a color document.