Sign In

lansing

User Group
Trusted Members
Join date
2-May-2017
Last activity
12-Feb-2018
Posts
30

Post History

Post
#1170455
Topic
Color matching and prediction: color correction tool v1.3 released!
Time

DrDre said:

lansing said:

DrDre said:

lansing said:

DrDre said:

lansing said:

Hi I ran into problem on color matching 2 color palettes, every color matched except that one on row 3 column 2, it always turn to pink on the result.

http://www.framecompare.com/image-compare/screenshotcomparison/0E1NNN8U

This is a discrete color matching problem. The color matching algorithm assumes images with a continuous color distributions with smooth color gradients, so it’s not perfectly suited for this sort of problem.

However, the explanation is in row 5 column 1, which in the source image has almost the same color as row 3 column 2. That color has to become a bright pink, and so row 3 column 2 also becomes pink.

So does that mean the problem can be resolved if I can sort the order of the color in the source image into one long smooth gradient before passing to the program?

It might be…

I isolated that two color and do the testing on the order, but the program didn’t run.

source:
https://i.imgur.com/58gYXRb.png

ref:
https://i.imgur.com/KmO21IJ.png

The algo matches color distributions. With two colors there is not much of a distribution, so I don’t think that will work.

I have found the reason for my problem, that’s because I had stacked way too many color adjustments into one color palette, trying to do the one LUT for all scenes thing. I have tried to manually alter each color with 3DLut Creator and I actually see the problem on the curve. The more adjustment points I have on the curve, the harder I can move them and the less accurate they are.

Post
#1169872
Topic
Color matching and prediction: color correction tool v1.3 released!
Time

DrDre said:

lansing said:

DrDre said:

lansing said:

Hi I ran into problem on color matching 2 color palettes, every color matched except that one on row 3 column 2, it always turn to pink on the result.

http://www.framecompare.com/image-compare/screenshotcomparison/0E1NNN8U

This is a discrete color matching problem. The color matching algorithm assumes images with a continuous color distributions with smooth color gradients, so it’s not perfectly suited for this sort of problem.

However, the explanation is in row 5 column 1, which in the source image has almost the same color as row 3 column 2. That color has to become a bright pink, and so row 3 column 2 also becomes pink.

So does that mean the problem can be resolved if I can sort the order of the color in the source image into one long smooth gradient before passing to the program?

It might be…

I isolated that two color and do the testing on the order, but the program didn’t run.

source:
https://i.imgur.com/58gYXRb.png

ref:
https://i.imgur.com/KmO21IJ.png

Post
#1169860
Topic
Color matching and prediction: color correction tool v1.3 released!
Time

DrDre said:

lansing said:

Hi I ran into problem on color matching 2 color palettes, every color matched except that one on row 3 column 2, it always turn to pink on the result.

http://www.framecompare.com/image-compare/screenshotcomparison/0E1NNN8U

This is a discrete color matching problem. The color matching algorithm assumes images with a continuous color distributions with smooth color gradients, so it’s not perfectly suited for this sort of problem.

However, the explanation is in row 5 column 1, which in the source image has almost the same color as row 3 column 2. That color has to become a bright pink, and so row 3 column 2 also becomes pink.

So does that mean the problem can be resolved if I can sort the order of the color in the source image into one long smooth gradient before passing to the program?

This post has been edited.

Post
#1155175
Topic
Color matching and prediction: color correction tool v1.3 released!
Time

DrDre said:

lansing said:

Williarob said:

Just create a second LUT and apply it after the first one. It should not affect the behaviour of the first LUT unless it is modifying the same color, and if that is the case you want it to modify that color anyway. LUTs can be stacked, it’s pretty neat, and really easy in After Effects of Resolve.

I tried it, I cropped out just the yellow block and did a color match, and then apply it on top of the first LUT in photoshop, but the whole image turned greenish.

target:

reference:

result after global LUT + shoe LUt:

There’s no guarantee it will work, because if you leave out one part of the frame, it may alter the unique colors of those parts as a conseqience of getting the correct yellow.

Looks like that’s the case for me. I redo the crop on the shoes with a bigger surrounding, the yellow was matched but it also affected the rest of the image.

Post
#1154742
Topic
Color matching and prediction: color correction tool v1.3 released!
Time

Williarob said:

Just create a second LUT and apply it after the first one. It should not affect the behaviour of the first LUT unless it is modifying the same color, and if that is the case you want it to modify that color anyway. LUTs can be stacked, it’s pretty neat, and really easy in After Effects of Resolve.

I tried it, I cropped out just the yellow block and did a color match, and then apply it on top of the first LUT in photoshop, but the whole image turned greenish.

target:

reference:

result after global LUT + shoe LUt:

This post has been edited.

Post
#1154684
Topic
Color matching and prediction: color correction tool v1.3 released!
Time

DrDre said:

lansing said:

DrDre said:

lansing said:

DrDre said:

lansing said:

hakkaibills93 said:

Dragon Ball Kai 1-98 use a differrent color palette than the dragon ball z series, all colors channel aren’t exactly the same that’s why you can’t duplicate z colors to them.
For 99 to the end you can for sure use as it was not done the same way

about dragon ball z bluray , colors are almost okay, film also have tint in the white and some slight alteration like the yellow being too much green and other slight things and yes there is crushed black, overbrightness and excessive saturation

movies also need color correction for what i have seen from the funi movies bluray

your references are good but for the kai 1-98 just keep the color exactly as they are

i already tried this tools and it only fail when the source you want to correct isn’t suppose to have the same colors (i tried with the dbz first episode preview that have goku arm in red (color error) that was corrected in the episode…i tried to color match the dbox footage with the broadcast colors and it didn’t change the arm color cause it wasn’t suppose to have this color

The issue I have with the program is that the algorithm tends to miss color in smaller areas of the image. This example shows it:
http://www.framecompare.com/image-compare/screenshotcomparison/0FCNNN8U

The big areas like the sky, the mountain and trees are all matched, but smaller areas like the wrist band, Goten’s robe, the rocks, Videl’s shoes are all missed.

I would advice using more color spaces, because it can definitely be much improved. Here a used 100 color spaces with the smoothing parameter set to 0.01:

It takes a little while to get some experience how to use the tool, but it should give satisfactory results most of the time. In general more color spaces will produce a more accurate result, but it will be slower. Increasing the smoothing parameter reduces artifacts, but also the accuracy, such that you often need more color spaces for a better result.

Um I am using the 64 bit version and it stated 250 color space instead of 100, and the result of the 250 color space is exactly the same as 10, looks like a bug there.

Yeah I have already improved my workflow to get better accuracy, such as auto-aligning and cropping the images with photoshop, which improve the accuracy by some noticeable margins. I just want to see if it can be improved on the color matching of those smaller areas, like the shoes and rocks.

The tool was never tested for 64 bit, so I can’t comment on that. Williarob compiled a 64 bit version as a favour, but beyond that it’s at this point not supported. Either way I got a better match with the settings I used for the 32 bit version.

How would you deal with those smaller areas?

There are three options:

  1. More color spaces
  2. A lower smoothing factor
  3. After performing a global correction, zoom in on the part you want to improve

What do you mean by zoom in? For example I want to match the yellow shoes after the first color match. I zoom in and do another color match just on the shoes, but then how do I stack the second match on top of the first?

Post
#1154483
Topic
Color matching and prediction: color correction tool v1.3 released!
Time

DrDre said:

lansing said:

DrDre said:

lansing said:

hakkaibills93 said:

Dragon Ball Kai 1-98 use a differrent color palette than the dragon ball z series, all colors channel aren’t exactly the same that’s why you can’t duplicate z colors to them.
For 99 to the end you can for sure use as it was not done the same way

about dragon ball z bluray , colors are almost okay, film also have tint in the white and some slight alteration like the yellow being too much green and other slight things and yes there is crushed black, overbrightness and excessive saturation

movies also need color correction for what i have seen from the funi movies bluray

your references are good but for the kai 1-98 just keep the color exactly as they are

i already tried this tools and it only fail when the source you want to correct isn’t suppose to have the same colors (i tried with the dbz first episode preview that have goku arm in red (color error) that was corrected in the episode…i tried to color match the dbox footage with the broadcast colors and it didn’t change the arm color cause it wasn’t suppose to have this color

The issue I have with the program is that the algorithm tends to miss color in smaller areas of the image. This example shows it:
http://www.framecompare.com/image-compare/screenshotcomparison/0FCNNN8U

The big areas like the sky, the mountain and trees are all matched, but smaller areas like the wrist band, Goten’s robe, the rocks, Videl’s shoes are all missed.

I would advice using more color spaces, because it can definitely be much improved. Here a used 100 color spaces with the smoothing parameter set to 0.01:

It takes a little while to get some experience how to use the tool, but it should give satisfactory results most of the time. In general more color spaces will produce a more accurate result, but it will be slower. Increasing the smoothing parameter reduces artifacts, but also the accuracy, such that you often need more color spaces for a better result.

Um I am using the 64 bit version and it stated 250 color space instead of 100, and the result of the 250 color space is exactly the same as 10, looks like a bug there.

Yeah I have already improved my workflow to get better accuracy, such as auto-aligning and cropping the images with photoshop, which improve the accuracy by some noticeable margins. I just want to see if it can be improved on the color matching of those smaller areas, like the shoes and rocks.

The tool was never tested for 64 bit, so I can’t comment on that. Williarob compiled a 64 bit version as a favour, but beyond that it’s at this point not supported. Either way I got a better match with the settings I used for the 32 bit version.

How would you deal with those smaller areas?

Post
#1154397
Topic
Color matching and prediction: color correction tool v1.3 released!
Time

DrDre said:

lansing said:

hakkaibills93 said:

Dragon Ball Kai 1-98 use a differrent color palette than the dragon ball z series, all colors channel aren’t exactly the same that’s why you can’t duplicate z colors to them.
For 99 to the end you can for sure use as it was not done the same way

about dragon ball z bluray , colors are almost okay, film also have tint in the white and some slight alteration like the yellow being too much green and other slight things and yes there is crushed black, overbrightness and excessive saturation

movies also need color correction for what i have seen from the funi movies bluray

your references are good but for the kai 1-98 just keep the color exactly as they are

i already tried this tools and it only fail when the source you want to correct isn’t suppose to have the same colors (i tried with the dbz first episode preview that have goku arm in red (color error) that was corrected in the episode…i tried to color match the dbox footage with the broadcast colors and it didn’t change the arm color cause it wasn’t suppose to have this color

The issue I have with the program is that the algorithm tends to miss color in smaller areas of the image. This example shows it:
http://www.framecompare.com/image-compare/screenshotcomparison/0FCNNN8U

The big areas like the sky, the mountain and trees are all matched, but smaller areas like the wrist band, Goten’s robe, the rocks, Videl’s shoes are all missed.

I would advice using more color spaces, because it can definitely be much improved. Here a used 100 color spaces with the smoothing parameter set to 0.01:

It takes a little while to get some experience how to use the tool, but it should give satisfactory results most of the time. In general more color spaces will produce a more accurate result, but it will be slower. Increasing the smoothing parameter reduces artifacts, but also the accuracy, such that you often need more color spaces for a better result.

Um I am using the 64 bit version and it stated 250 color space instead of 100, and the result of the 250 color space is exactly the same as 10, looks like a bug there.

Yeah I have already improved my workflow to get better accuracy, such as auto-aligning and cropping the images with photoshop, which improve the accuracy by some noticeable margins. I just want to see if it can be improved on the color matching of those smaller areas, like the shoes and rocks.

Post
#1154371
Topic
Color matching and prediction: color correction tool v1.3 released!
Time

hakkaibills93 said:

you have effectively recover dark things but you now less see things in the white things as you can’t almost not see anymore the details i was talking before
that’s why your references have to be “perfect” else you’ll loose something if you don’t have perfect reference you’ll have to alter manually colors and work with tone yourself to recover the dark things without destroying the white

That is what the program was supposed to do. It successfully matched the color of the star war blu-ray to the reference. The whites in the matched image was blown out because the white in the reference image was also blown out. Your observation is really off.

From my tests, it was more like the priority decision of the color matching has to do with the size of the color. If I have a big chunk of red in the image, that red will have higher matching priority than some smaller reds. For example, in my comparison images, if I make a montage reference by stacking this current shot and a close up shot of the yellow shoes and then use that resulting LUT on this current shot, the yellow shoe would be matched.

Post
#1154261
Topic
Color matching and prediction: color correction tool v1.3 released!
Time

hakkaibills93 said:

but anyway it’s mean to be wrong, funimation blue ray have dark colors cause it have crush black and overbright so if you “copy” colors to the kai footage which don’t have crush black you’ll have some colors too dark than others
you have to think that (cause it’s not totally accurate like i said with some green in the yellow for example) if you have a dark blue with a crush black footage, your goal is to get the same blue as bright as it was crushed black so a middle blue instead of the dark blue

What? I don’t think that’s how the algorithm of the program works.

Post
#1153697
Topic
Color matching and prediction: color correction tool v1.3 released!
Time

hakkaibills93 said:

Dragon Ball Kai 1-98 use a differrent color palette than the dragon ball z series, all colors channel aren’t exactly the same that’s why you can’t duplicate z colors to them.
For 99 to the end you can for sure use as it was not done the same way

about dragon ball z bluray , colors are almost okay, film also have tint in the white and some slight alteration like the yellow being too much green and other slight things and yes there is crushed black, overbrightness and excessive saturation

movies also need color correction for what i have seen from the funi movies bluray

your references are good but for the kai 1-98 just keep the color exactly as they are

i already tried this tools and it only fail when the source you want to correct isn’t suppose to have the same colors (i tried with the dbz first episode preview that have goku arm in red (color error) that was corrected in the episode…i tried to color match the dbox footage with the broadcast colors and it didn’t change the arm color cause it wasn’t suppose to have this color

The issue I have with the program is that the algorithm tends to miss color in smaller areas of the image. This example shows it:
http://www.framecompare.com/image-compare/screenshotcomparison/0FCNNN8U

The big areas like the sky, the mountain and trees are all matched, but smaller areas like the wrist band, Goten’s robe, the rocks, Videl’s shoes are all missed.

Post
#1144137
Topic
Color matching and prediction: color correction tool v1.3 released!
Time

clutchins said:

lansing said:

animefan said:
Your wrong about that everything dragon ball related needs color correcting i saw the color problems in all of it.

The Funimation BD releases are color accurate

…accurate to what exactly? What is your master color reference?

The dragon ball z power level card by Bandai that came out in the early 90s:
https://www.youtube.com/watch?v=9p-a-4oAnsw

This post has been edited.

Post
#1143185
Topic
Color matching and prediction: color correction tool v1.3 released!
Time

I found bugs in the 64 bit version, the program didn’t run often on the first double click. I saw it flashes in the task manager and went away. I have to do a second double click to open it.

There’s also a bug with the ui. When I’m loading in the images, if I don’t do file->close on those image preview windows, the images may not be loaded, causing the button to be grayed out or still stuck with the last successful loaded image.

Post
#1142969
Topic
Color matching and prediction: color correction tool v1.3 released!
Time

animefan said:
Your wrong about that everything dragon ball related needs color correcting i saw the color problems in all of it.

The Funimation BD releases are color accurate, they only have the over saturation and over brightening problem, in which they toned them down in the Buu saga release. That’s why it’s worth color matching them to the kai version.

Post
#1142777
Topic
Color matching and prediction: color correction tool v1.3 released!
Time

animefan said:

Are you workinh on to release a color edit of this the 13 movies,the two specials and the original dragon ball? Also will it include the broadcast audio?

My plan is to color match the kai version to funimation season bluray/level version, and then put all the color matched LUTs of all the episodes into an library so everyone who has a copy of kai can download and reproduce them on their own. No pirating.

I have made williarob’s montage approach into a semi-auto script so it can be scaled up to a community project which everyone can contribute (nobody can do it alone). However that approach took a big blow from these test results. Those new color palette used for Kai (at least before the Buu saga) is screwing up the color matching. For example, in the reference image, the skin color of Krilin and Vegeta are different, but on the kai, they are the same. So the color matched result came out turning them into just one color. The test results also show that a montage for just one scene is not good enough (well at least for scene with no motion), sampling frames from similar scenes may give better result, as shown on the second comparison.

Even though this is a good news that kai is still matchable, everything has to go back to the drawing broad again because I have no idea how to scale that up now.

As for the 13 blu-ray movies, there’s really nothing you need to do except lowering the saturation and highlight.

Post
#1142460
Topic
Color matching and prediction: color correction tool v1.3 released!
Time

I finished writing a montage building script in Vapoursynth that mimic what Williarob did in his tutorial video. It will takes a scene change text file as input and create montage base on scene changes.

https://forum.doom9.org/showthread.php?t=174796

But then when I took it to further test on my Dragon Ball Z and I found out that montage that was build from one single scene in anime doesn’t do well in color matching…

target:

reference:

match:

A total fail, none of the color matched except the dark blue…

However, when I create a montage base from 3 similar scenes, the result is surprisingly better:

target

reference

match

The only color that was noticeably off is red.

This post has been edited.

Post
#1103645
Topic
Color matching and prediction: color correction tool v1.3 released!
Time

Wow I was just starting to roll out my plan for the Dragon Ball Z color correction project like 12 hours ago in the kanzenshuu forum and this bomb DROPS. Huge thanks!

The program extracted and installed successfully, only problem is that it didn’t delete the MCRInstaller.exe file after the installation.

Tested on two 8k montage matching, works without a problem, RAM usage peak at 3.1G

Post
#1102210
Topic
Color matching and prediction: color correction tool v1.3 released!
Time

Sad for Windows users. Hopefully a Windows user with a 64 bit Matlab would come by…soon.

And I have another problem that I couldn’t think of a solution, even though I match using a montage, still some time there’re a couple of color that were off, but when I cropped and only match those problem areas, the color did matched. How do I fix these?

Post
#1100634
Topic
Color matching and prediction: color correction tool v1.3 released!
Time

So I did a result comparison between LUTs created by the original 1080p size, a shrank 720p, and by using the “fast processing mode” in the program on my anime. And I saw noticeable differences in the light areas on the intended “faster” methods compare to the LUT created from the original size. So in my conclusion, don’t resize or use the “fast processing mode” if you’re going for quality, because the visual differences are not neglectable.

Post
#1096441
Topic
Color matching and prediction: color correction tool v1.3 released!
Time

DrDre said:

lansing said:

Williarob said:

alexp120 said:

alexp120 said:
If you look at williarob’s “Restoring Color to Red Faded Film” video

http://thestarwarstrilogy.com/starwars/post/2016/09/12/Restoring-Color-to-a-Faded-Eastman-Print-of-Star-Wars

…beginning at the 15:47 mark, he explains how you can build a model containing a wide variety of colors. You are grabbing sample frames of the shot that you are working on from both the reference video and the test video, and creating LUT’s of all the colors in that shot.

I noticed that the number of sample frames of the shot from both the reference video and the test video is 16. Why 16?

I think it had to do with how big the montage image was going to be. I didn’t want any blank areas, so it had to be 4x4 or 3x3 or 5x5, but if I recall correctly, any bigger than 4x4 at that resolution (each frame about 848x360? - something like that) would just leave the color match tool hanging indefinitely on my system. Your system may be more or less powerful than mine, so your mileage may vary.

If you were using the tool to grade a film like Raiders or Star Wars then I would recommend you take 8 to 16 frames from every scene (rather than shot) and turn them into montages, one of the source (or test) frames and one of the target (reference) frames and create a LUT for each scene.

In my testing, I did try grabbing up to 64 small frames (smaller than SD) from an entire reel of Star Wars and generating montages and a single LUT for an entire reel, but found it was much less accurate than doing it on a scene by scene basis. However, this was probably because the reel did not have consistent colors (parts of the reel were scanned with different color settings) so 1 LUT to rule them all was not possible until all the shots had been color balanced first.

However you may be able to create a single montage for an entire episode of what you are working on, especially if the color changes are consistent.

This is a very late follow up because I was slacking off. I created a montage reference image and it sure is more accurate. My question now is is 4x4 montage the overall best size as a reference for a scene? Or is it better to build the montage with 1 frame from each second in the scene?

My second question is what is the minimum resolution we can use to create the montage without affecting quality when building the color matching model in the program? It sounds too crazy to build the montage with the full size 1920x1080 frames.

Montages will work, but the more frames the more difficult it is to match the frames, with a higher probability of artifacts. For the fast mode the frame size is significantly reduced.

For making a montage, what is a good size for any individual frame to reduce to? Because it makes sense to reduce a 4x4 montage of 1920x1080 frames, but it doesn’t make sense to reduce the size of a 4x4 montage that is 1920x1080, since each frame inside the montage is very small already.

Post
#1096425
Topic
Color matching and prediction: color correction tool v1.3 released!
Time

Williarob said:

alexp120 said:

alexp120 said:
If you look at williarob’s “Restoring Color to Red Faded Film” video

http://thestarwarstrilogy.com/starwars/post/2016/09/12/Restoring-Color-to-a-Faded-Eastman-Print-of-Star-Wars

…beginning at the 15:47 mark, he explains how you can build a model containing a wide variety of colors. You are grabbing sample frames of the shot that you are working on from both the reference video and the test video, and creating LUT’s of all the colors in that shot.

I noticed that the number of sample frames of the shot from both the reference video and the test video is 16. Why 16?

I think it had to do with how big the montage image was going to be. I didn’t want any blank areas, so it had to be 4x4 or 3x3 or 5x5, but if I recall correctly, any bigger than 4x4 at that resolution (each frame about 848x360? - something like that) would just leave the color match tool hanging indefinitely on my system. Your system may be more or less powerful than mine, so your mileage may vary.

If you were using the tool to grade a film like Raiders or Star Wars then I would recommend you take 8 to 16 frames from every scene (rather than shot) and turn them into montages, one of the source (or test) frames and one of the target (reference) frames and create a LUT for each scene.

In my testing, I did try grabbing up to 64 small frames (smaller than SD) from an entire reel of Star Wars and generating montages and a single LUT for an entire reel, but found it was much less accurate than doing it on a scene by scene basis. However, this was probably because the reel did not have consistent colors (parts of the reel were scanned with different color settings) so 1 LUT to rule them all was not possible until all the shots had been color balanced first.

However you may be able to create a single montage for an entire episode of what you are working on, especially if the color changes are consistent.

This is a very late follow up because I was slacking off. I created a montage reference image and it sure is more accurate. My question now is is 4x4 montage the overall best size as a reference for a scene? Or is it better to build the montage with 1 frame from each second in the scene?

My second question is what is the minimum resolution we can use to create the montage without affecting quality when building the color matching model in the program? It sounds too crazy to build the montage with the full size 1920x1080 frames.

To the top