I just needed to share my recent experience. My client (OK, actually my sister and Brother-in-Law) asked me to restore an image from a 40 year-old, halftoned newpaper clipping, lovingly preserved under clear contact paper.
Though I’ve had great success removing moiré patterning, PhotoAI simply could not deal with a low resolution newpaper halftone B&W image (and probably some additional moiré due to introduction of artifacts due to linear array scanning).
I then applied 3rd party FFT, and some additional smoothing algorithm to remove as much excessive digital noise as possible without too much losing image definition, but while PhotoAI is great at face reconstruction, the overall restoration was still lacking.
I then decided to applied a 3rd party A.I. Colorization to the denoised image. Then a reapplication of PhotoAI to this newly colorized version and WOW! My “client” immediately called me on the phone after email receipt and was totally blown away by the restoration (which I said was “good enough!” LOL). The results are attached!
Thanks again.
All due to prepping the original which I believe is a better match to the A.I. training data. Now consider incorporating A.I. Colorization for B&W images which definitely provided a much improved image restoration!
