scanning film with sd quattro

Artemka wrote:

A multilayer sensor do not give in principle higher resolution. If you scan B/W you must get rid of the Demosaic. This should give you the same resolution as all RGB pixels are avaible w/O interpolation. In stead you must adjust the 2x2 color channels for minimum derivation. This depend on the light source while making a scan/picture with the Bayer cam. It should be best a flash light because of vibration and spectral distribution.

I think RT have the capability to adjust color channels w/o Demosaic

I've tried processing with RPP (Raw Photo Processor), which claims it can take the Luminosity without Demosaic-ing it. The results are somehow inferior to using normally processed files in dcraw. When enlarged to 400% the RPP file shows jagged lines and blotches and is very inconsistent in quality. As I mentioned, I can sort of overcome it by stitching and ending up with higher res, but man, every time I see the micro detail of Foveon...

The issue in the flow w/o Demosaic ist that the luminace response form the RGB is different. This gives the noise. If the three, assuming 2 equal green sensitivities, channels are equalized the noise vanish.

Could you please write your workflow to make a conversion this way?

The only way I tested some b/w shots was via dcraw. There is a command line option to generate un-demosaic linear tiff from raw. You get then a gray linear image with full resolution but with the filter effects of the bayer filter in front of the sensor,

Then you have to equalize the gray 2x2 underlying pattern which is the result of your exposure light source and the dye filters. To construct the equlization you have to exposure a master image with about 18%. This master image must be inverted so that you get 0.18/image_value. The inverted master image have be multiplied with every tiff to get the color pattern out of the gray image.

The link below describe the usage for a custom demosaic but the construction is similar.

http://therefractedlight.blogspot.de/2014/04/cook-your-own-raw-files-part-3-demosaic.html

The difference is that the author construct a bayer pattern image where you need an inverted equalization image. The second step by the author is the custom demosaic which is not needed for b/w. The layer operation is done PS but RT is also possible.

An issue is that hot pixels where not removed up to now. Possible that dcraw have an option to interpolate hot pixels and not demosaic at the same time. This what I remember from more than 2 years ago.