scale/re-size images without blurring of line quality

[D.T. Nethery]

A long standing request from many users over the years has been to improve the way TVPaint handles images that have been scaled or rotated with the Transform tool or Perspective (Warp) tool -- the lines in TVPaint get noticeably blurry after applying only a couple of transformations , in a way that does not happen in some other bitmap drawing apps , such as Photoshop , for one example. I realize it has been pointed out before that transforming a drawing in Photoshop also results in some line quality degradation (blurriness) , that it is simply the nature of working with pixels , I understand that ... but frankly, I do not notice it as much in Photoshop or Krita or Clip Studio Paint as in TVPaint. I've been able to transform (rotate and scale) drawings in Photoshop multiple times with very little noticeable degradation of the image. The image degradation of transformed drawings is much more noticeable in TVPaint.

Therefore, I am making this feature/improvement request again. I hope this request may be near the top of the developer's "To-Do" list. I am reminded about it because of the impending release of Niels Krogh Mortensen's Animation Paper app (formerly known as PAP - "Plastic Animation Paper") which mentions this specifically in his update on the beta release:

"We recently finished lots of advanced X-Sheet functionality, reference layers for video/rotoscoping or 3D reference, copy/paste of connected drawings (clones), off-the-pegs feature, etc, etc. Animation Paper has our awesome line drawing engine, with great advantages, like lines don't degrade when copied, moved or scaled. Layers are fully in place. Light-table (onion skinning) is good. Everything works super fast and realtime - on both PC and Mac"

https://animationpaper.com/2019/03/27/e ... breakfast/

[NathanOtano]

When I tested it regarding photoshop, I found the same result for the same transformations at the same resolution (translating, rotating, scaling, and combining multiple ones). I also heard that tvpaint and adobe have kind of the same algorythm seems legit to me.
Personnally I sometimes use "none" as a setting instead of "smart" cause I find it beter looking at the end, even with the aliasing. Fabrice told me back in the days that the "best" mode is essentially for photos.

Here are the main reasons why photoshop seems to have a better algorythm :
- You usually work in much greater canvas in photoshop than in tvpaint. 1080 and A4 300dpi doesn't have the same pixels to work with...
- Because of the resolution you really see the pixels and work with it, that doesn't happen in photoshop (and mayeb we work more zoomed in in animation
- In animation (and because of the tvpaint destructive workflow) you usually transform multiple times in a row, degrading further the pixels. That doesn't happen in photoshop.
- Those algorythms work well on photos, cause there is a dependancy beetween areas and you have to keep your global coherence (so "extend" pixels accordingly). When working with lines, you want to have the same sharpness on the sides of the strokes. But it's not really possible, there is no informations when you transform the pixels, the standard algorythm for photos just doesn't want to create new informations cause it's not it's purpose (but hey there is different algorythm choices in photoshop so why not). You need a new algorythm specifically for lines, but then it really depends on the brush you're using...

Regarding PAP if it's a brand new algorythm specifically designed for lines it makes sense to me I think tvpaint's algorythm is quiet standard and quiet old right now.

[D.T. Nethery]

When I tested it regarding photoshop, I found the same result for the same transformations at the same resolution (translating, rotating, scaling, and combining multiple ones). I also heard that tvpaint and adobe have kind of the same algorythm seems legit to me.
Personnally I sometimes use "none" as a setting instead of "smart" cause I find it beter looking at the end, even with the aliasing. Fabrice told me back in the days that the "best" mode is essentially for photos.

Here are the main reasons why photoshop seems to have a better algorythm :
- You usually work in much greater canvas in photoshop than in tvpaint. 1080 and A4 300dpi doesn't have the same pixels to work with...
- Because of the resolution you really see the pixels and work with it, that doesn't happen in photoshop (and mayeb we work more zoomed in in animation
- In animation (and because of the tvpaint destructive workflow) you usually transform multiple times in a row, degrading further the pixels. That doesn't happen in photoshop.
- Those algorythms work well on photos, cause there is a dependancy beetween areas and you have to keep your global coherence (so "extend" pixels accordingly). When working with lines, you want to have the same sharpness on the sides of the strokes. But it's not really possible, there is no informations when you transform the pixels, the standard algorythm for photos just doesn't want to create new informations cause it's not it's purpose (but hey there is different algorythm choices in photoshop so why not). You need a new algorythm specifically for lines, but then it really depends on the brush you're using...

Regarding PAP if it's a brand new algorythm specifically designed for lines it makes sense to me I think tvpaint's algorythm is quiet standard and quiet old right now.

I'll try the "none" setting instead of "smart" .

Your explanation about how the algorithms are intended to work is helpful and I'm sure you are correct , but to my eyes the transform in TVPaint does seem to noticeably blur the images more quickly than in Photoshop (and to clarify: I was speaking of line art , not photos) .

"You usually work in much greater canvas in photoshop than in tvpaint. 1080 and A4 300dpi doesn't have the same pixels to work with..."

I'm working at 4K resolution in TVPaint (to downscale to 2K for final output) and I still seem to notice the blurry line quality after only a few transformations with either the Transform Tool or Perspective tool . I have become used to it over the years , so after I've transformed a frame many times I'll either re-draw it on a new layer and then merge it back into my animation layer or if it is only minor blurriness I will use the Curves FX to sharpen the line ... it doesn't matter when doing rough animation , but I'm speaking of clean-up line work , where small adjustments are made to parts of the drawings ... however, this announcement of the feature in the new version of PAP which specifies "lines don't degrade when copied, moved or scaled" caught my eye immediately and reminded me about this long-standing request , so I thought I would mention it again.

"Regarding PAP if it's a brand new algorythm specifically designed for lines it makes sense to me . I think tvpaint's algorythm is quiet standard and quiet old right now."

Yes , I suppose it is standard, but if TVPaint's current algorithm is standard (based on code written many years ago) then my improvement request is : please develop a new algorithm
for Transforming which is specifically designed for line art to retain the maximum line quality.

.

[NathanOtano]

Yeah we both talk about line art. I agree that a new could be a nice improvement!
You also can try medium, it's what it's made for. I often see myself switching from one algortyhm to the next while transforming, picking the best option (which is quiet heavy in the process i agree)

(BTW You should really test yourself the same line art at the same resolution, I assure you it's the same result in photoshop)

[schwarzgrau]

I could't find the post, but I'm pretty sure somewhere here in the forum someone from the TVPaint team mentioned, that Photoshop actually has some superior transforming, cause even if it's not a smart layer it handles layers as kind of vector layers, with a bitmap mapped to it (which doesn't make sense anymore, when writing it down). But probably it's the same algorithm, but used in a more smart architecture.

[Cardin]

Hi,

I like the new library idea, however, I wish you could copy image to layer from reference library. Seems like a no-brainer. Not as anim brush, as it is now, that is just so unintuitive if you have big images etc.

Another thing is that when I use an image as a guide from the library the reference image is always on top. Even with opacity set lower, the pixels I draw blend underneath the reference image.

So maybe you could add a switch to the guide settings so that the layer you draw on is on top?

Right now Im going back to my previous workflow: importing in reference images into layers. So hopefully you can fix this

Or am I missing some setting?

TVPaint Staff, this user is a spam bot. Its mining old posts and posting as new so as to appear authentic...

Reference..

http://forum.tvpaint.com/viewtopic.php?f=21&t=9629

Hi,

I like the new library idea, however, I wish you could copy image to layer from reference library. Seems like a no-brainer. Not as anim brush, as it is now, that is just so unintuitive if you have big images etc.

Another thing is that when I use an image as a guide from the library the reference image is always on top. Even with opacity set lower, the pixels I draw blend underneath the reference image.

So maybe you could add a switch to the guide settings so that the layer you draw on is on top?

Right now I'm going back to my previous workflow: importing in reference images into layers. So hopefully you can fix this

Or am I missing some setting?

Please delete my post after you've deleted theirs.

[Anim-Illustrate]

Hello.

Interesting post.

Just comming out of a big production this was a pretty big issue for us.

Animation characters in different sizes in a scene could easily become a nightmare when(working in the ruff stages) using all the transform tools.

The characters would simply get to damaged.

Sometimes we would scale up up the specific character animated in big size and then scale down inside the scene. And sometimes that would be a guessing game if the characters would be placed in the right place in a scene.

You start to hear whispers in the corners "vector/Toonboom".

It could be so awesome if TvPaint could work on getting away from blurry line quality

Thanks.