Danbooru

Scan Quality

Posted under General

Lately, well not just lately, there has been a rather huge amount of relatively low-quality scans being uploaded.
These scans are often either have mass amounts of scan_artifacts, scanlines, creases or moire textures.

As someone who actively cleans scans, this is quite annoying to see.

  • scan_artifacts can often be cleaned with no effort. There is always the case where effort is required , but that is not saying they can't be cleaned.
  • scanlines often cannot be cleaned without damaging the image. This set is a nice example. The only image there that I would say could even possibly be cleaned would be post #1238294, and even then it would kill the quality of the background to do so.
  • moire textures, the worst of the lot. This depends on how bad it is, but most often cannot be cleaned without ruining the quality of the scan.
  • creases really shouldn't be uploaded any circumstances. Scans from posts like post #1256043, can often be decreased no problem. Scans from books on the other hand cannot be without a bit of redrawing, as the image is two seperate pages.
  • There is also scans like this where the contrast is obviously too low.
  • Aswell as the odd image which it is a single page of a two page scan (Like post #1240069 or post #1238315).

I might just be being picky, but this isn't a scan site. The only scans that should be posted are ones that are cleaned and complete.
If it's something that can be cleaned, it should be. If it can't then it shouldn't be uploaded.
If any of these were digital images, they would probably just get flagged and deleted for the quality.
I'm all for having more content on the site, but should this really include poor quality scans?

So I have two questions.

  • Does the content outweigh the quality of the scan (So if it includes any of the above, should it be uploaded)?
  • Should we start pushing all new scans into the queue regardless of the uploader, so that any contrib+ scans get checked? (This could be done with a bot if need be)

Updated by jxh2154

DakuTree said:

  • Does the content outweigh the quality of the scan (So if it includes any of the above, should it be uploaded)?

In my opinion, yes, the content does outweigh the quality of scan. If the content is really good, then I do not consider if the quality of the scan is less than perfect. Only if the scan quality is significantly bad does it get me to weigh against approval of an image I'd otherwise like. If a better scan of an image is available or possible, uploaders should always be encouraged to contribute it and parent it.

EB said:
If a better scan of an image is available or possible, uploaders should always be encouraged to contribute it and parent it.

This really shouldn't even be needed though. The first version should be the only version that should exist.
As I see it, if an active user posts a poor quality scan then they are being lazy. There should be no need for additional versions of the same scan to be posted, nor any need to keep the uncleaned versions around.

Single pages of two page-scans should really not be approved/uploaded even if the quality is good.
Hell, the reason I posted this in the first place is because I knew some lazy uploader would post these four images without even putting effort into cleaning or attempting to put the two pages together.

I knew some lazy uploader would post these four images without even putting effort into cleaning or attempting to put the two pages together.

I know you're more or less aiming that at me (and one other user), but you're kind of assuming that most people know how to do stuff like that.

I have absolutely no skills or knowledge in cleaning/drawing/stitching/vectoring/etc. or any software and a high-end computer to do that.

I just find things and upload them...

If you wish to expunge them though, feel free.

Updated

Jigsy said:
...

I'm not really aiming at you, I just knew someone would post it considering the content.

And you don't even need any skills or a high-end computer to clean images.
A simple run with either Greycstoration or Neat Image (or both in some cases) removes most artifacts, then applying clone tool / spot healing brush to things that didn't get picked up.

DakuTree said:
As I see it, if an active user posts a poor quality scan then they are being lazy. There should be no need for additional versions of the same scan to be posted, nor any need to keep the uncleaned versions around.

It must have occurred to you that, much like uploaders aren't necessarily skilled in comprehension of Japanese (hence the romanization help threads etc.), they aren't necessarily skilled in image editing either.

In fact now it's just one more quality I can envy you for. I know for one that I would post more scans I encounter if I was able to clean them. But if they look decent enough as is, like EB said it's worth posting, especially if it's unlikely that someone else will post a better version.

Cyberia-Mix said:
It must have occurred to you that, much like uploaders aren't necessarily skilled in comprehension of Japanese (hence the romanization help threads etc.), they aren't necessarily skilled in image editing either.

It's true I might be expecting too much, but it still doesn't change the fact the poor quality scans are being uploaded when they could be cleaned first.

I know for one that I would post more scans I encounter if I was able to clean them.

Just an idea. What about a simple thread for people to post scans that could be cleaned? Either that or a howto:clean_scans wiki.

DakuTree said:
Single pages of two page-scans should really not be approved/uploaded even if the quality is good.

I disagree. I they work as standalone images, they should be fine. Only element I see extending into the other image in this case is Momo's veil and tail, which is not a big deal in my opinion.

Doesn't look like these images can just be simply pasted together. Without having access to the source material, what can be done in this case?

EB said:
Doesn't look like these images can just be simply pasted together. Without having access to the source material, what can be done in this case?

Well considering the content of the scan, I really doubt that will be the only scan of it. There will be several people who scan it.

The bit missing is on the side of the book, this and this looks like they are on a dust jacket rather than the actual book, meaning the missing bit could be scanned.
Thankfully there is nothing on each version of the image that changes on the missing area, meaning it could just be copied over to the alternate versions of the images.

Although that does mean waiting a bit.

Cyberia-Mix said:
But if they look decent enough as is, like EB said it's worth posting, especially if it's unlikely that someone else will post a better version.

I suppose others differ on this, but this goes for both scans and online sample images for me: if the quality of the illustration is really worthwhile, I think it should be uploaded. Something is better than nothing when it comes to enjoyable images, and I can't always expect someone will come along and upload a "perfect" version later on. There gets to a point where artifacts, low resolution, etc. will drown out the quality of a good illustration, but it really needs to be judged on a case by case basis. I do think uploaders should try to find the best version possible from the start, but some (like me) lack access to source material and the ability or desire to fool with image editing software in the case of scans.

I'm also of the opinion that, though cleaning a scan is probably a good idea, but if not done well, photo-manipulation can also easily ruin an image (*cough* *cough* ). Most people cannot be assumed to have the skills required, and in that case having an image at all is better than not, so long as it's of sufficient quality to be approved. We could potentially have a tag (or a thread as EB suggests) to request someone with sufficient skill do a clean-up. The cleaned version could then be posted under the cleaner's account, and parented appropriately.

This would mirror our policy when it comes to better quality or higher resolution images being found and posted to replace earlier worse-quality images.

In short, having a good quality image in an unrefined form is better than not having it at all since otherwise many wouldn't be aware of its existence. Likewise, the unrefined form is likely better than an unskilled editor botching an image in an attempt to prepare it for upload. If you have the skill, yes, you should probably make use of it; if you don't, it's probably better to post the raw and ask for help.

EDIT: If you do post a lot of bad scans though, keep in mind it may not reflect positively on you for promotion purposes. If you work with a lot of scans, it might be worth looking into and training yourself to clean them properly.

Updated

Shinjidude said:
We could potentially have a tag (or a thread as EB suggests) to request someone with sufficient skill do a clean-up.

I think this would be the best idea.
A thread may work better, as people could post there before they upload, aswell as having a place to double-check the quality before uploading.

A tag would be useful too, but I honestly think we need to work on getting people to actually tag scan first. There is who knows how many images that lack it.

D'Eye said:
If I'm not mistaken there's a tag for this: fixme.

That tag is from people copy-pasting tags from yande.re.

DakuTree said:
Just an idea. What about a simple thread for people to post scans that could be cleaned? Either that or a howto:clean_scans wiki.

Or both, so it's not just you working for everyone. But I support this anyway. I'd gladly learn to do it myself too if it's easy enough.

EB said:
I suppose others differ on this, but this goes for both scans and online sample images for me: if the quality of the illustration is really worthwhile, I think it should be uploaded. Something is better than nothing when it comes to enjoyable images, and I can't always expect someone will come along and upload a "perfect" version later on.

A bit off-topic, but do you have some debatable examples of the latter at hand?

Updated

Cyberia-Mix said:
Yeah online image samples questionable for upload.

First thing to come to mind for me is santa_matsuri, since her artwork is amazing though some of the samples I've seen in the past are of such awful image quality that's its hard to consider uploading or approving. I wasn't remembering it exactly, but it turns out the only deleted image here (post #962969) serves as an apt example. That would normally be a must-approve for me even with significant defects in image quality, but that's too much... On the other hand, several others that were uploaded are more borderline and I think are worth having.

Updated

Of course if they get that far, there's really no saving them. With the amount of work it'd take to make look good, you'd essentially just be creating your own derivative_work. Even if you did redraw the image and it looked decent, you shouldn't post it by our guidelines, since you've become the artist yourself to a large degree.

I was thinking more of things along the lines of what Dakutree pointed out in his OP post. Things that can be relatively easily done semi-automatically, or with minimal touching up.

I agree that some degree of artifacting can be acceptable. Especially for images where the artifacting is in the original uploaded by the artist (more common that you might think, especially with earlier works). User-based clean-up should only be used in the case where the artifacting is caused by users (or 3rd parties) and not the artist (e.g. via scan, or for images that have been copy-saved in multiple generations and the originals are no longer available).

Updated

I know. I was going off into a tangent anyway since I wasn't talking about scans there. Since the image is a sample of a doujin cover, it could potentially have been scanned of course (and likely would be an approvable image in that case), but that requires someone willing to go after the source material. The unlikeliness of such things happening would be one thing motivating me to approve it if it were more borderline (for instance, if it was that same size but only had lightly noticeable artifacts).

The issue of scan quality is fairly simple in my mind:

1. Users should not be expected to modify the images whatsoever before posting.

2. Users should however upload the best quality version of an image that they are aware of.

3. If they or another user come across a better version later, upload and parent.

Straightforward, I'd say.

1