Comments on post #2824323

@Mikaeri on post #2824323 ( original and 1 more drawn by furisuku)

2017-08-17T19:59:23-04:00

CodeKyuubi said:

Honestly, I do tag lossy-lossless where I find it in my day-to-day browsing, which doesn't happen very often with digital images, and with scans it's a more questionable usage as a scan by nature has image noise due to it being a physical-to-digital conversion. I feel that the very presence of the scan tag should basically equate to a user that the image has some form of artifacting, unless overly processed.

I don't know why you feel that I'm trying to say this whole time that my version is superior, as through my silence I have conceded that your image has slightly more image data and not reversed the parent/child relationship again. As I said before, I see your point of view, henceforth I have kept the relationship as it is. All I wanted to say was that this post deserves the lossy-lossless tag for the past several posts.

My perception was that you were trying to justify making yours the parent because this does suffer from a defective quality -- that it being lossy-lossless is enough to move the parenting around. But if you mean to say that this should stay the parent and also have it tagged, then by all means (of which I've now done).

But I'm kind of a foulmouthed fool sometimes, so, y'no... Apologies.

And yeah, the consideration of lossy-lossy has not escaped me once it came to my attention. Should all scans be uploaded as png, or should they be uploaded as the filetype of the original, or always as a jpg. In my opinion the data loss of a maximum-quality jpg is, as you said, minimal. Twitter is a weird thing. It's a lossy format, for sure, but that is just the nature of how twitter works. I definitely have tagged many twitter images with jpeg artifacts if it's severe, particularly in images with lots of red color data, but then it begs the question if we start tagging images with lossy-lossy, then do we start tagging images with no quality loss from upload as lossless-lossy? It's a dilemma, and there's probably no answer that will satisfy everyone. I wouldn't mind tagging every twitter image as lossy-lossy, though the problem there is the fact that most users don't do tag aftercare after migrating tags from a twitter post, and it would quickly bloat pixiv images with the tag as well.

Pretty much. The existence of jpeg artifacts is definitely subjective, and the tagging noise that would result from having a tag such as lossy-lossy would be too much technicality to keep up with.

Further reason as to why I disagree with part of the existence of a lossy-lossless tag because in the first place we shouldn't be uploading those sorts of images, and it's typically reserved in cases where a third-party has uploaded a "better" image out of innocence from a third-party imageboard (e-shuushuu's pointless pngs, for example), and thus should be flagged. The current policy is that uploading such images should be discouraged except in cases where nothing, really, could be done better. A JPG would've been fine, but then there's that problem of introducing noise... on an already noisy image, albeit.

@CodeKyuubi on post #2824323 ( original and 1 more drawn by furisuku)

2017-08-17T19:12:00-04:00

Mikaeri said:

If the definition is to be trusted as such, then lossy-lossless should be much more heavily populated, and lossy-lossy should be considered.

I'm usually not a fan of stitches, which is why I rarely do them, but if you're going to stitch two uploads and claim that while making both simultaneously worse and not doing anything to improve them, that's in poor taste in my eyes. Hell, I would've likely upped this earlier had I known you be stingy enough to do this. Yes, these scans are "raw" and "littered with artifacts". My problem is that you're selectively tagging images you don't like to try to make them look worse, while not doing the rest for others.

By the bare definition, yes, this is lossy-lossless, but there is frankly no other way to simply save it as one whole in the same lossy format without introducing further artifacting. And just because that is, doesn't make yours better. In fact, as I've said multiple times before, it's worse.

The question, then, is how exactly, or as to what legitimacy does this 3rd party edit have over yours, or what yours has over this one?

The last time I opened up a discussion on the forum about its usage it simply ended at "tag it what it is". I don't disagree with the existence of the tag nor if it belongs here (because it obviously does) but I do disagree as to how it's being attributed, and what exactly does it mean when a post is considered "worse" than others simply because of a quality of being taxing on filesize and not from the pure image itself. And going by the current definition, lossy-lossy doesn't share the same usage as jpeg artifacts, unless you want to consider mass-tagging everything uploaded from twitter as lossy-lossy.

EDIT: To add, our TL group exports in jpg simply because it's an easier format to work with (especially on most web browsers), and the final export (with 100% quality, with 2x2 1x1 1x1 subsampling), the filesize doesn't go up nearly as much as it would in PNG. Artifacting is negligible enough for what it is.

But why I don't do that here is because I care about the preservation of the image even if it is unnecessarily bloated, because that's simply how I do 3rd party edits. If this is the insistence and the length we're going to, then maybe I should just convert this to JPG and do an in-place replacement, so we can end this here.

Honestly, I do tag lossy-lossless where I find it in my day-to-day browsing, which doesn't happen very often with digital images, and with scans it's a more questionable usage as a scan by nature has image noise due to it being a physical-to-digital conversion. I feel that the very presence of the scan tag should basically equate to a user that the image has some form of artifacting, unless overly processed.

I don't know why you feel that I'm trying to say this whole time that my version is superior, as through my silence I have conceded that your image has slightly more image data and not reversed the parent/child relationship again. As I said before, I see your point of view, henceforth I have kept the relationship as it is. All I wanted to say was that this post deserves the lossy-lossless tag for the past several posts.

And yeah, the consideration of lossy-lossy has not escaped me once it came to my attention. Should all scans be uploaded as png, or should they be uploaded as the filetype of the original, or always as a jpg. In my opinion the data loss of a maximum-quality jpg is, as you said, minimal. Twitter is a weird thing. It's a lossy format, for sure, but that is just the nature of how twitter works. I definitely have tagged many twitter images with jpeg artifacts if it's severe, particularly in images with lots of red color data, but then it begs the question if we start tagging images with lossy-lossy, then do we start tagging images with no quality loss from upload as lossless-lossy? It's a dilemma, and there's probably no answer that will satisfy everyone. I wouldn't mind tagging every twitter image as lossy-lossy, though the problem there is the fact that most users don't do tag aftercare after migrating tags from a twitter post, and it would quickly bloat pixiv images with the tag as well.

@Mikaeri on post #2824323 ( original and 1 more drawn by furisuku)

2017-08-17T18:37:16-04:00

CodeKyuubi said:

And so images should be, if the artifacts are clearly visible through the naked eye without zooming in. In this case the artist applied noise to the image, as some are wont to do, which helps deal with gradient banding, which at the same time disguises the artifacts, which are also image noise.

I already explained my reservations of tagging a by-definition lossy-lossless image without jpg artifacts as one, as the main criteria to tag lossy-lossless on Danbooru is to have the artifacts from having previously been a compressed jpg file. I don't see why you're arguing this point, if you have a problem with the usage of lossy-lossless not being the actual true-definition of lossy-lossless, you should bring it up on the forum, but we have a lot of similar cases where Danbooru usage differs from real-world usage.

If the definition is to be trusted as such, then lossy-lossless should be much more heavily populated, and lossy-lossy should be considered.

I'm usually not a fan of stitches, which is why I rarely do them, but if you're going to stitch two uploads and claim that while making both simultaneously worse and not doing anything to improve them, that's in poor taste in my eyes. Hell, I would've likely upped this earlier had I known you be stingy enough to do this. Yes, these scans are "raw" and "littered with artifacts". My problem is that you're selectively tagging images you don't like to try to make them look worse, while not doing the rest for others.

By the bare definition, yes, this is lossy-lossless, but there is frankly no other way to simply save it as one whole in the same lossy format without introducing further artifacting. And just because that is, doesn't make yours better. In fact, as I've said multiple times before, it's worse.

The question, then, is how exactly, or as to what legitimacy does this 3rd party edit have over yours, or what yours has over this one?

The last time I opened up a discussion on the forum about its usage it simply ended at "tag it what it is". I don't disagree with the existence of the tag nor if it belongs here (because it obviously does) but I do disagree as to how it's being attributed, and what exactly does it mean when a post is considered "worse" than others simply because of a quality of being taxing on filesize and not from the pure image itself. And going by the current definition, lossy-lossy doesn't share the same usage as jpeg artifacts, unless you want to consider mass-tagging everything uploaded from twitter as lossy-lossy.

EDIT: To add, our TL group exports in jpg simply because it's an easier format to work with (especially on most web browsers), and the final export (with 100% quality, with 2x2 1x1 1x1 subsampling), the filesize doesn't go up nearly as much as it would in PNG. Artifacting is negligible enough for what it is.

But why I don't do that here is because I care about the preservation of the image even if it is unnecessarily bloated, because that's simply how I do 3rd party edits. If this is the insistence and the length we're going to, then maybe I should just convert this to JPG and do an in-place replacement, so we can end this here.

@CodeKyuubi on post #2824323 ( original and 1 more drawn by furisuku)

2017-08-17T17:48:24-04:00

Mikaeri said:

And I disagree with its usage. If you tag it here, then your post deserves jpeg artifacts along with the non-stitched images.

If you tag this lossy-lossless, then so should post #2804604 and many other posts. You have done nothing but provide a third-party edit of an image; I simply provided a better one, one which loses absolutely zero data.

And so images should be, if the artifacts are clearly visible through the naked eye without zooming in. In this case the artist applied noise to the image, as some are wont to do, which helps deal with gradient banding, which at the same time disguises the artifacts, which are also image noise.

I already explained my reservations of tagging a by-definition lossy-lossless image without jpg artifacts as one, as the main criteria to tag lossy-lossless on Danbooru is to have the artifacts from having previously been a compressed jpg file. I don't see why you're arguing this point, if you have a problem with the usage of lossy-lossless not being the actual true-definition of lossy-lossless, you should bring it up on the forum, but we have a lot of similar cases where Danbooru usage differs from real-world usage.

@Mikaeri on post #2824323 ( original and 1 more drawn by furisuku)

2017-08-17T17:19:47-04:00

CodeKyuubi said:

Edit: My opposition to legitimate lossy-lossless images is that it is deceptive to the end-user, giving them the impression that an image is higher quality than it actually is. I see you removed the tag, but I would recommend re-applying it as this image very much fits the definition of Danbooru's usage of said tag.

And I disagree with its usage. If you tag it here, then your post deserves jpeg artifacts along with the non-stitched images.

If you tag this lossy-lossless, then so should post #2804604 and many other posts. You have done nothing but provide a third-party edit of an image; I simply provided a better one, one which loses absolutely zero data.

@CodeKyuubi on post #2824323 ( original and 1 more drawn by furisuku)

2017-08-17T16:56:30-04:00

Mikaeri said:

It came natively in a digital format. I thought the same at first, which is why I held off for a second, but in the case where it came as such then the only thing to do is to preserve it as such.

The other rips (post #2824285, post #2824284) don't come like this (or to such a heavy degree of the artifacting as you say), so presumably we should assume good faith, that the source is as correct as the publisher made it. If there is visible "artifacting" then that was probably intended by the artist, or was simply a result of a scan made digital (when the image wasn't drawn on a digital medium to begin with).

I did the comparison and checked the pixel values -- you did not do any sort of restoration to it (no redraws, no color restoration, etc), so yours is worse as it is lossy-lossy. Lossy-lossless isn't the answer, sure enough, but if you're going to do that you're going to also have to tag post #2804604 and others like it as such.

Restoring scans... that's fine. In your ballpark. But as said above, you should either do at least this much or forego it.

Yes, I did say that if I did it properly, which I did not as I've since forgone fixing digital images because of the issues of some people uploading the unmodified, lower-quality raws as parents. I do photoshop work to give the end users the best result I can, which is rendered meaningless when it is childed because no one looks at the children except the people who impulsively favorite everything.

That said, no it is not artist intended. For novel illustrations, this can be seen because purchasing the book gives you access to clean versions of the image, as opposed to the artifacts from illustration rips. It is third-party modified by the novel distributor in order to encourage purchasing the books. For digital magazines, I find they can vary from distributor to distributor, as some give out 1920px images with very minimal artifacting, while some only give 1200px or 1500px with heavy artifacting, so it's likely that they are also third-party modified to disincentivize digital rips, whether legitimately purchased or not.

And as for the pixiv images, it's been in my mind. By definition it belongs to the lossy-lossless category, but common usage for Danbooru is when visible artifacts from over-compression are present.

Edit: My opposition to legitimate lossy-lossless images is that it is deceptive to the end-user, giving them the impression that an image is higher quality than it actually is. I see you removed the tag, but I would recommend re-applying it as this image very much fits the definition of Danbooru's usage of said tag.

@Mikaeri on post #2824323 ( original and 1 more drawn by furisuku)

2017-08-17T16:24:14-04:00

CodeKyuubi said:

If I would have done it properly, I would have cleaned the image before saving it at max quality. The issue then is that someone comes in, uploads the raw image littered with artifacts, and parents the clean image to it. So, I've given up cleaning digital images and will only clean scans.

I can understand your point of view, but I feel a lossy-lossless is not the way to go around it. That's why I linked the source if people want to grab the raws for themselves.

It came natively in a digital format. I thought the same at first, which is why I held off for a second, but in the case where it came as such then the only thing to do is to preserve it as such.

The other rips (post #2824285, post #2824284) don't come like this (or to such a heavy degree of the artifacting as you say), so presumably we should assume good faith, that the source is as correct as the publisher made it. If there is visible "artifacting" then that was probably intended by the artist, or was simply a result of a scan made digital (when the image wasn't drawn on a digital medium to begin with).

I did the comparison and checked the pixel values -- you did not do any sort of restoration to it (no redraws, no color restoration, etc), so yours is worse as it is lossy-lossy. Lossy-lossless isn't the answer, sure enough, but if you're going to do that you're going to also have to tag post #2804604 and others like it as such.

Restoring scans... that's fine. In your ballpark. But as said above, you should either do at least this much or forego it.

@CodeKyuubi on post #2824323 ( original and 1 more drawn by furisuku)

2017-08-17T16:01:58-04:00

Mikaeri said:

Because you're introducing a second round of lossy compression when you save it in JPG. That is simply how it works. If you're going to stitch it, at least do this much.

If I would have done it properly, I would have cleaned the image before saving it at max quality. The issue then is that someone comes in, uploads the raw image littered with artifacts, and parents the clean image to it. So, I've given up cleaning digital images and will only clean scans.

I can understand your point of view, but I feel a lossy-lossless is not the way to go around it. That's why I linked the source if people want to grab the raws for themselves.

@Mikaeri on post #2824323 ( original and 1 more drawn by furisuku)

2017-08-17T15:58:32-04:00

CodeKyuubi said:

The source image is lossy and rife with artifacts, why would you re-save it into a lossless format?

Because you're introducing a second round of lossy compression when you save it in JPG. That is simply how it works. If you're going to stitch it, at least do this much.

EDIT: As with post #2804604

@CodeKyuubi on post #2824323 ( original and 1 more drawn by furisuku)

2017-08-17T15:44:24-04:00

Mikaeri said:

Lossless stitch

diff on post #2824317

The source image is lossy and rife with artifacts, why would you re-save it into a lossless format?

@Mikaeri on post #2824323 ( original and 1 more drawn by furisuku)

2017-08-17T15:30:52-04:00

Lossless stitch

diff on post #2824317