After the initial results with only one color channel looked good I wanted to see what happens when I read all 3 color channels and put them in a picture.
First the electronics had to upgraded and all color channels connected to the arduino. Then the software got a complete rewrite, since managing the hacked together code from the first examples got more and more complicated. The file format got changed as well from a text file to a binary file. This reduces overhead a lot (previous Image Size at 1/4 resolution was 73MB and that is only grayscale! With RGB colors I now get the same image size)
So after a lot of coding and a lot of testing and a lot of weird errors that happen when you try to process binar data I got to the point where I could get a first color-test scan.
I used my phones LCD as a light source, made a slow scan and changed the color while scanning. This was looking very promising.
To increase image quality and to make testing a lot easier I sealed the sensor part of the scanner with black foam board and black tape. There were a lot of problems with weird sensor readings and I suspected that the t-shirt I used to hang over the scanner didn't make a very good seal.
The ribbon cable feedthrough was also sealed with black foam that is usually used for light seals in analog camera repairs.
This is the first color scan. It shows a couch and a White/Gray/Red blanket without any post processing except adjusting dark/light values. The results had basically no color vibrance and incorrect white balance. I think this was a bad Idea, since the sensor doesn't have an IR filter and the room was lit with a halogen lamp.
Scans at sunset provided similar results, but were still very 'flat' and show little depth in color. Also the trees were moving, which is not very good for a scanning camera.
I tried to use lightroom to get the image as close to 'normal' as possible. It was very difficult since the green trees and blue sky have much too similar color. The small orange square on the crane is actually an illuminated red sign and it was a very nice sunset with a colorful sky.
The results from the test with my phone as light source were very good, but real world scans are incredible bad, which is discouraging. I will probably make some tests with better controlled lighting to see if the missing IR filter is actually a problem or the bad color response has a different cause.