A Scan-Digitizing Algorithm of the Printed Original Based on Use of Near-Neutral Scales

摘要

Scan-digitizing is one of the main ways to digitize the original. Building scanner profiles are usual practice to get the consistent color. In the paper, we describe how to get the best possible 'Raw RGB' image from a scanner. First we measured the accuracy of those different input profiles. The profile accuracy depends on the relationship between scanners RGB and the corresponding LAB or XYZ values. Then we compared the range of those input profile tones. The ideal tone response curve can be decided by setting the scanner gamma. So a good input profile can faithfully render the dark and murky appearance of an underexposed image or the washed-out appearance of an overexposed one. In order to get a good scanner profile we proposed a method to determine a suitable tone response curve. A new algorithm was then developed according to N-component LUT-based input profile. Experiments were carried out to compare these scanner profiles. The results showed that the ICC profile generated by the new algorithm has obvious advantages in accuracy and grey predicting.