Infrared Light Source

[spaceChRiS]

I was pondering about such a project for a long time already, and now I found your repository which goes 80 % the direction I wanted to go. I definitely need to build such a light source. However, what bothers me, as I have a lot of scratched and dirty negatives (and positives), is if and how one could add an infrared light source for detection of these scratches and dust.

Some thoughts on infrared, maybe this is interesting or somebody reading this might even have some answers …

• Which wavelength would be suitable for scratch detection? I was not able to find what wavelength the fuji or noritsu scanners are using. If this information is not available, that's maybe something to test with different LEDs.
  • Infrared focus: If the wavelength is too far off of the visible spectrum, the result might be too blurry, but a little bit of blur might be acceptable for the IR channel.
• Infrared diffuser: Will the material combination you chose also work for infrared?
• Infrared power: As most cameras have a filter for infrared light, it may be useful to have enough power for reaching enough SNR despite having the IR cut filter on the sensor. Another option would be IR conversion of the camera.

[spaceChRiS]

For transparency: I've also asked this question partially on pixls.us: https://discuss.pixls.us/t/wavelength-of-infrared-channel-for-scratch-dust-detection/47941.

[raphaelb74]

@spaceChRiS I completely agree with your approach. I'm currently in the process of rebuilding my setup as well, using a Sony A7R III that will be converted to monochrome with the IR filter removed. I'm quite confident it would be possible to develop something that replicates the functionality of the old "ICE" algorithm.

@jackw01 would you think it'll be possible to add space for IR LED ? Also do you plan on making a larger PCB layout for 120 film ? v3 perhaps ?

Below the board from Fuji SP3000 RGB+IR:

[jackw01]

Done - see https://github.com/jackw01/scanlight/blob/main/big.md

The IR LEDs used in my design are 850nm - the most common and most efficient wavelength and the typical wavelength for security cameras and other illumination applications. I am fairly confident this is the same wavelength used by professional film scanners, or close enough. An IR-converted camera is required to see light at this wavelength. Based on my testing with an IR-converted webcam, the dye layers of the film block a little of the light, producing a very low contrast image (I had assumed they would be completely transparent to 850nm, but apparently not), and any dust on the film shows up as black spots which could easily be isolated with a simple thresholding algorithm.