I didn't actually do any color correction, only levels. That's basically the color that was present in the original scan, and thus I guess close to the color present on the original page? Assuming the original scan was well done. Honestly, it'd probably be best to rescan that image and use the descreening built into the scanner driver (which always seems to do a better job of it than any attempts to descreen after the fact). At the very least, the "full original scan" needs to be far, far less compressed IMO. Here is a detail of the left wing, JPG artifacts out the wazoo.
I think the bigger concern here is the moire effect, but that's more of an indication that the scan isn't going to get much better, since it's picking up how the ink lies on the page. It's asking a lot - a good scan of potentially dodgy printed off-camera screenshot. The fact it needs to be colour corrected is a clue it might not be doable. But who knows.
Agreed. I was honestly surprised the scanlines became visible. You can just barely see a regular pattern when looking at the halftones, but nothing that suggests that the scanlines would even be recoverable. Ideally, the chosen resolution would be the sensor's resolution. We don't really need any horizontal upsampling.
I can try scanning it again. For the JPG compression, I used 95% quality through Paint.Net. Any suggestions with what dpi would be ideal? I can also apply a descreening filter while scanning if that helps(as mentioned). I'll try to have a new scan up this weekend after looking at suggestions. Edit: Would PNG be more preferable?
PNG is going to be massive in file size, but there won't be any compression artifacts. I'd recommend PNG.
Agreed. File size is no issue to me, so the new scan that I will post will be in PNG format. I appreciate the input from you guys because I want to do this screenshot justice considering what it is. Thanks again for the input! />
Because you're using that extra storage space to preserve noise. Photos and scans are what lossy formats like JPEG were designed for.
Did some measuring and came up with this as a pixel scaled version: Could easily be off by one or 2 pixels one way or another, as the image scanned isn't correctly proportioned (it's stretched vertically)
Lossy formats like JPEG sacrifice noise data with respect to human perception. We can't be sure that they won't remove signal data that the additional processing being done in this thread could make use of.
More correctly, you're seeing a combination of the native pixel aspect ratio of an NTSC Megadrive, as modified by the geometry of the display device that was photographed. If you scale to the exact pixel height of the raw graphic, this scan should be narrower.
If you wanted the definitive super duper bestest scan ever possible, you'd scan it hundreds of times and then average out all of the resulting images. You know, something completely unrealistic. The Retro policy on the wikis is to prefer JPEG for scans, because with PNGs you're preserving the imperfections at the expense of huge file sizes. Also many of the tricks behind PNG optimisation start failing at that level (not to mention it would take an age to process). That being said, I always use 100% on the Paint.NET JPEG scale so compression artifacts are as hidden as possible. Though do remember, we still hope to find the ROM of this some day - if and when that happens, the pixels will be preserved. Fun fact: We actually have this exact sort of redundancy on Sonic Retro right now: These cropped up magazine screenshots were uploaded to the internet in the early 2000s. But we have the Sonic 2 8-bit rolling demo ROM now, so we don't actually need these blurry things anymore. They'd have about the same amount of use to us whether they were PNG or JPEG - resolution matters more. But but but but but We could really do with a full scan of the whole book, because that hasn't been done.
Is that scanned at more than 300 dpi? I believe that's the very minimum resolution when doing magazines.
