Why Colour Negative is Orange
Why?
Still the problem of approximately reconstructing the illuminant from dye densities appears to be very difficult for some reason. Why?
Good. At least this is possible.ARRI built a scanner for motion picture film which is uses a custom designed light source and sensor, and they have very good color scientists to make it work.
the fact that different emulsions have very different characteristics, and change somewhat due to production, storage and processing, makes things very tricky
the main reason why color negative film is more difficult than reversal film, is that we have to cancel out the masking dyes and raise the contrast a lot, which makes any problems from a light-film-sensor-processing missmatch much more obvious.
That must be the root cause of all our trouble. Aerocolor to the rescue!
Agree. I assume the motion picture industry has to deal with the same problems. They would probably purchase film from the same batch and process it together.
That must be the root cause of all our trouble. Aerocolor to the rescue!
yes, quality control is much tighter in the motion picture industry. film is fresh and in cool storage, lab have (or at least used to have) special technicians that ran chemical test and control strips to keep the chemistry constant, so that the shots of different days could be intercut without color shifts (this was specially important in days when films were printed analog).
they also have fewer and more uniform film stocks than the photo world (which is actually a downside in terms of choice).
The $200.000 scanner doesn't cost $200.000 because it has an out of this world light source or sensor or insanely good inversion routine, but because it's freaking fast and at the same time mechanically insanely precise at placing the film during the caption.
Of course, $200.000 can buy you a decent tricolour+ir light and a monochrome sensor, too.
the mechanical part is no that hard to solve, pin registered film transport has been around for many decades.
its first version actually did only about 1 frame per second at highest resolution (which is only about 4000ppi).
what you pay for is for R&D of people with good wages and precision manufacturing in a high cost country, combined with low volume sales to make up for it.
interestingly, they use a temperature stabilised sensor and light source that is self-calibrating to achieve consistent results, which is an indication that they take color rendering quite serious.
Not to mention that they'll likely solve frame-to-frame alignment at least partly in the digital domain. That's a lot easier, more flexible and cheaper than fixing it entirely in hardware. It's similar to the 'steady shot' algorithms that have been around for 25 years in the video domain.
$200.000 can buy you a decent tricolour+ir light and a monochrome sensor
but to get good colours out of it you need some seriously experienced engineers.
I also prefer to scan as positive/color slide and then do the inversion manually using curves.
After you write the post on RA-4 papers could you please make a tutorial on curve-based inversion?
Brilliant! Thanks a lot. I will definitely try it. Your procedure looks fairly straightforward to automate. I never tried scanning a negative with a film scanner as a slide so not quite sure what to expect, but the first histogram in your post looks somewhat strange. I assume the peaks at the right are from the light source shining through the sprocket holes. If this is the case can you improve the dynamic range by cropping or covering them?Hope it helps in any way!
If this is the case can you improve the dynamic range by cropping or covering them?
I scan with a digital camera. I found that I can extend the dynamic range of the scan by increasing the color temperature of the light source to the maximum (more blue) and then using custom white balance on the camera to compensate for the orange mask. I set the exposure such that the mask is scanned as very light grey (1 to 1/3 stop below clipping).
Remember, if you are going by the histogram on live view screen, that is not a true representation of raw sensor data that you will be working with.
Your procedure looks fairly straightforward to automate.
the first histogram in your post looks somewhat strange. I assume the peaks at the right are from the light source shining through the sprocket holes.
There is probably a typo in the sentence starting with "I can then now the information in the histogram".
it should have a cheap stand alone tool, with a pay to update every few years
We use cookies and similar technologies for the following purposes:
Do you accept cookies and these technologies?
We use cookies and similar technologies for the following purposes:
Do you accept cookies and these technologies?