Efficient classification of scanned media using spatial statistics

We address the automatic classification of scanned input media in order to improve color calibration. Since scanner responses vary significantly according to the type of input, a media dependent color calibration for a scanner is desirable for accurately mapping scanner responses to a standard color space. To assist such media dependent calibration, we propose an efficient algorithm for automated classification of input media into four major classes corresponding to photographic, lithographic, xerographic, and Inkjet. Our technique exploits the strong correlation between the type of input medium and the spatial statistics of corresponding images, which may be observed in the scanned images. Adopting two spatial statistical measures of dispersion and periodicity, and utilizing extensive training data, we determine well separated decision regions to classify the input medium with a high confidence level. Experimental results over an independent test data set validate the results.