Example 4. Counterfeit Detection Based on Ink Color
In this example, we use pseudocolor to highlight subtle differences in RGB color that might be encountered in certain types of imaging, including remote sensing, pathology, microscopy, and printing. Here, we apply this technology to the color printing of a US $100 bill.
Below, please see the unretouched flat bed scan (4.1) of a portion of a USA $100 note at 300dpi.
Next, the same region is shown (4.2) with the color of the top half of the image slightly increased in the green channel (constant luminosity). This change is undetectable to the human eye. This simulates a 'counterfeit' note wherein the color of the top half of the $100 bill is too green.
From the second image, three random points are picked. A new image is generated using the color values of these randomly picked pixels for new R, G, B axis set to full scale red, green, and blue. All the other pixels in the image are color adjusted to this new color axis by orthonormalization.
Each of following sequential images (below, 4.3) differs only by the orthonormalization, that is in turn based on the specific pick of three random pixels. Most of the images show a clear demarcation along the length of the note corresponding to the 'authentic' bottom half and the 'counterfeit' top half. Nothing has been assumed about the nature of the 'counterfeiting', so this method should detect any type of subtle color change including ink spectra, ink blending into the paper, and the color of the paper itself.
Higher resolution data is shown next as an animated gif (4.4):
Source code in .nb and .PDF formats.