Not sure if I understand this correctly. I guess the scanner only does one side illumination, and scanning from the other side. I don’t think it has two ccds.
I did a comparison of different inpainting algorithms in this post:
The whole technique I used is described a couple of posts before.
As only the missing parts of the image are inpainted, there is no overall reduction in sharpness. Furthermore, there’s at least one inpainting method for which I would attest that the result is not distinguishable from regular image content. Fair enough, my example does not show scratches at areas such as faces where the content is more crucial, but from my experience with old family photographs, the results are really excellent and the manual effort is close to zero, only some initial mask adjustment (which is necessary only one time per film stock).