Serious glass sheet optical anatomy

by adrien

g161

The first result of the algorithm

1im

A

1back

B

glass-spectrums

C

A & B & C : the raw material, pictures and glass transmission spectra (blue for different densities of the sheet, black for a reference white glass of the same factory)

*

In the previous post I showed some pictures produced by my technique of background removal for making “as accurate as possible” photos of glass sheets.

The photos were taken on a TL back-lighted, which is highly irregular. What you have to do is basically to multiply the picture A by the inverse of the picture B, after calibrating the pictures A & B together by making sure that their background has the same lightness. Of course you should do this in RAW linear 16bit!

This simple operation removes: the optical artefacts (the lens of the camera filters more light in the picture’s borders) the electronic ones (the CCD is slightly non-linear). It also automatically generates a “perfect” white balance, whatever your light source is… The only pity is that it adds some noise, but for our application this is fine.

*

I developed a method for selecting parts of the glass sheets based on their spectral transmission, by combining spectrophotometric measurements, math, and the corrected pictures :

g161-png_deltay_14-0

The 3 zones A, B, C have an average difference of 14% in luminance when Blue-sky light passes through them.

other glass sheets characterized in the same way:

rose_or_clair-png_deltay_10-0

3275xx_2225-png_deltay_10-0