Color detail issues in Fujifilm video files

[tupp]

1 hour ago, Attila Bakos said:

Okay, for those who find this kind of stuff interesting, here is a comparison of Cr channels (with added contrast for easy visualisation) from the C70, X-T3, and X-H2s

Thanks for comparison!

The Fuji cameras definitely introduce a significant blotchiness that is not inherent in the Canon footage.

It would be interesting to see unaltered footage without the added contrast.  I wish that the Canon position/framing was aligned more closely with the Fujis.

[androidlad]

4 hours ago, Attila Bakos said:

Okay, for those who find this kind of stuff interesting, here is a comparison of Cr channels (with added contrast for easy visualisation) from the C70, X-T3, and X-H2s:
https://www.dropbox.com/s/pbfjdx4c2zxhhzq/chroma_compare.mp4?dl=1

C70 is C-Log3, 100Mbps
X-T3 is F-Log, 400Mbps
X-H2s is F-Log, 360Mbps

There is definitely an improvement over the X-T3.
(I only received this files from someone, so I can't do other comparisons.)

Thanks. There seems to be very little, if any, difference between X-H2s and X-T3 in this clip.

[Attila Bakos]

24 minutes ago, androidlad said:

Thanks. There seems to be very little, if any, difference between X-H2s and X-T3 in this clip.

The X-H2s is better, less blocky, but I was expecting a bit more as well. Maybe ProRes internal is better, but the guy who provided the clips doesn't have a CF express card so we couldn't test that. When he gets one he'll send me another comparisons. What I could test is ProRes RAW HQ external, and it's really good, but it's 15GB for one minute.

[Jay60p]

Months ago I did extensive testing comparing my X-T3 color to my Canon T2i.  I finally came to two
conclusions, different from my last post.

1) I find the Cr channel completely useless for evaluating the chroma resolution or any other color quality of the X-T3. It is quite bizarre and completely misleading. Instead, study the usual RGB channels which appear completely normal and are as detailed as the actual full color photo. There is no red resolution being lost in the Red RGB like there appears to be in the Cr channel. I’m guessing the Cr channel is useless for image manipulation too, which may be why it was so hard to find an App which would display it.

 

2) The Canon had always showed more saturation in the darker reds compared to the X-T3. So I went and studied the shot comparisons between cameras at the Imaging Resource Comparometer:

https://www.imaging-resource.com/IMCOMP/COMPS01.HTM

Select the “Still Life 400” samples for X-T3 and any other camera to compare the same scene.
In the X-T3 samples, The dark reds in the cloth on the left side appear lightened and/or slightly desaturated compared to all the other brands, not just Canon. This happens in solid color fields, not just red spots like foliage.

Why does the X-T3 lighten/desaturate dark reds? I’m guessing the face is the most important image for Fuji’s color science. Perhaps the Fuji engineers discovered that this could downplay skin blemishes, which are often of a darker red color than the rest of the skin.

In any case the good news is this:
I checked Fuji’s online specifications for the X-H2S.  And guess what, there is a provision for a setting
Labeled “Smooth Skin Effect”, listed right after “Color chrome Blue : Strong, Weak, Off”.
The actual settings are blank at the moment for the smooth skin effect. In future firmware there may
be an “off” setting available.  Maybe then those faint red leaves will appear in the foliage.

https://fujifilm-x.com/global/products/cameras/x-h2s/specifications/

[androidlad]

8 hours ago, Jay60p said:

Months ago I did extensive testing comparing my X-T3 color to my Canon T2i.  I finally came to two
conclusions, different from my last post.

1) I find the Cr channel completely useless for evaluating the chroma resolution or any other color quality of the X-T3. It is quite bizarre and completely misleading. Instead, study the usual RGB channels which appear completely normal and are as detailed as the actual full color photo. There is no red resolution being lost in the Red RGB like there appears to be in the Cr channel. I’m guessing the Cr channel is useless for image manipulation too, which may be why it was so hard to find an App which would display it.

 

2) The Canon had always showed more saturation in the darker reds compared to the X-T3. So I went and studied the shot comparisons between cameras at the Imaging Resource Comparometer:

https://www.imaging-resource.com/IMCOMP/COMPS01.HTM

Select the “Still Life 400” samples for X-T3 and any other camera to compare the same scene.
In the X-T3 samples, The dark reds in the cloth on the left side appear lightened and/or slightly desaturated compared to all the other brands, not just Canon. This happens in solid color fields, not just red spots like foliage.

Why does the X-T3 lighten/desaturate dark reds? I’m guessing the face is the most important image for Fuji’s color science. Perhaps the Fuji engineers discovered that this could downplay skin blemishes, which are often of a darker red color than the rest of the skin.

In any case the good news is this:
I checked Fuji’s online specifications for the X-H2S.  And guess what, there is a provision for a setting
Labeled “Smooth Skin Effect”, listed right after “Color chrome Blue : Strong, Weak, Off”.
The actual settings are blank at the moment for the smooth skin effect. In future firmware there may
be an “off” setting available.  Maybe then those faint red leaves will appear in the foliage.

https://fujifilm-x.com/global/products/cameras/x-h2s/specifications/

Great finding. But the "Smooth Skin Effect" is a blur effect only, it was added to GFX cameras a long time ago, it does not affect colour. Default is off and it has to be switched on.

[Jay60p]

4 hours ago, androidlad said:

Great finding. But the "Smooth Skin Effect" is a blur effect only, it was added to GFX cameras a long time ago, it does not affect colour. Default is off and it has to be switched on.

In that case, perhaps a custom LUT could adjust the Fuji dark reds to a darker richer red,
closer to the Canon look.
Beyond my abilities unfortunately.

[tupp]

6 hours ago, androidlad said:

the "Smooth Skin Effect" is a blur effect only, it was added to GFX cameras a long time ago, it does not affect colour.

Blurring definitely affects color:

Note that there are none of the more saturated tones in the blurred version.

Likewise, lowering resolution (within the same bit depth) reduces color depth.

 

[Attila Bakos]

17 hours ago, Jay60p said:

1) I find the Cr channel completely useless for evaluating the chroma resolution or any other color quality of the X-T3. It is quite bizarre and completely misleading. Instead, study the usual RGB channels which appear completely normal and are as detailed as the actual full color photo. There is no red resolution being lost in the Red RGB like there appears to be in the Cr channel. I’m guessing the Cr channel is useless for image manipulation too, which may be why it was so hard to find an App which would display it.

Quite the opposite, this is what my video is about. There is no RGB in the original file, only YCbCr. RGB is constructed from YCbCr. The Red channel you see in Resolve/Premiere/etc is detailed because it is constructed from Y and Cr. Since Y is full resolution and very detailed, the resulting Red channel will be detailed as well, even though Cr is heavily processed and has less resolution. So if you want to know why your image lost colors in detailed, colorful areas -even though you see a detailed red channel- then you have to go back to the original channels and examine them. And that is Y, Cb, and Cr.

[webrunner5]

The difference between YCbCr and RGB is that YCbCr represents color as brightness and two color difference signals, while RGB represents color as red, green and blue. In YCbCr, the Y is the brightness (luma), Cb is blue minus luma (B-Y) and Cr is red minus luma (R-Y)

When I was in the ENG world RGB was hardly ever mentioned. 

[Django]

On 7/16/2022 at 5:15 AM, Jay60p said:

2) The Canon had always showed more saturation in the darker reds compared to the X-T3. So I went and studied the shot comparisons between cameras at the Imaging Resource Comparometer:

https://www.imaging-resource.com/IMCOMP/COMPS01.HTM

Select the “Still Life 400” samples for X-T3 and any other camera to compare the same scene.
In the X-T3 samples, The dark reds in the cloth on the left side appear lightened and/or slightly desaturated compared to all the other brands, not just Canon. This happens in solid color fields, not just red spots like foliage.

Why does the X-T3 lighten/desaturate dark reds? I’m guessing the face is the most important image for Fuji’s color science. Perhaps the Fuji engineers discovered that this could downplay skin blemishes, which are often of a darker red color than the rest of the skin.

That's inherent to all Fuji X-Trans cameras. I remember it drove me crazy back when I had XT2 & 5DIII. 

The dark reds could never be as rich as on Canon. Some shades of red were almost pinkish red and I could never get them right in post without messing up the whole CS. I remember doing multiple side by side tests and the difference was night & day. The deep reds was by far the most annoying thing about Fuji color science to me..

Hopefully ProRes/ProRes RAW mitigates this on XH2S.

 

[Jay60p]

21 hours ago, Attila Bakos said:

Quite the opposite, this is what my video is about. There is no RGB in the original file, only YCbCr. RGB is constructed from YCbCr. The Red channel you see in Resolve/Premiere/etc is detailed because it is constructed from Y and Cr. Since Y is full resolution and very detailed, the resulting Red channel will be detailed as well, even though Cr is heavily processed and has less resolution. So if you want to know why your image lost colors in detailed, colorful areas -even though you see a detailed red channel- then you have to go back to the original channels and examine them. And that is Y, Cb, and Cr.

Okay, RGB may be constructed from YCbCr, but I don’t find reds only affected in small detail areas, they are equally affected in large color fields. I spent too much time shooting tests with tiny red details, assuming the reds were being muted by chroma noise reduction as suggested in other posts. The chroma noise reduction is not to blame, in fact I found it extremely effective in the Fuji Cr without blurring detail.

I studied Fuji jpg stills (also originally YCbCr?) in the app
GIMP (the only GUI app that would extract YCbCr components that I could find).

The contrast in the Fuji Cr component extract, displayed in B&W, was much lower than in the Canon Cr, giving it that misty washed-out look compared to the Canon Cr, but everything was still visible.
There was no detail blurring, edges were sharp. No red items missing compared to the Canon.

In fact, for green screen I much prefer the noise reduction In the Fuji. The edges of objects against a green screen were sharp and extremely clean, compared to much more noise in the green areas of the Canon Cr component.
Of course, the Canon T2i is an older, cheaper, and lower resolution camera so I am not too surprised.
But lost or blurred details in the Fuji Cr, I could not find.

Only lower red saturation in the full color shot.

And I also looked at Fuji RAW stills. I saw the dark reds were still lighter than the Canon and looked pretty much the same in red detail resolution & saturation as the jpegs. There were slight differences in the RAW between two different apps, just to add to the confusion!

The Fuji video Cr files you have shown - of the wooded areas with the red leaves - appear very cloudy and nearly devoid of any detail, which is what appeared so bizarre & useless to me. Probably because they were nearly monochromatic compared to my test shots.
We are not looking at the same types of files, and we're using different apps - You are using video and I am using jpeg & raw stills, for whatever that is worth. My jpeg Cr’s definitely show detail, So I’ll admit the Cr was not completely useless.

If there is loss of red detail specific to video, I can suggest comparing a Fuji jpeg still to a video frame Of the exact same shot with lots of small multi-colored items, but I am done and I leave that for someone else to enjoy!

[Attila Bakos]

@Jay60p This issue is mostly about video, and at this point no further tests are needed, the color loss is there, and the reason is known. And to be honest I don't think we'll get a fix for this, not for the X-T3/X-T4. I'm still not sure about the X-H2s, it seems to be better but I only received a few, not really ideal samples.
And I will say this again, 99% of the Fuji shooters won't ever notice this, so I don't want to take away the joy of shooting with a Fuji camera, but the issue is valid nonetheless.

[Jay60p]

Postscript:
My wife tried reading my last post.
She didn’t get very far.
She said we’re all crazy.
No doubt, She’s right.

[BTM_Pix]

55 minutes ago, Jay60p said:

My wife tried reading my last post.
She didn’t get very far.
She said we’re all crazy.
No doubt, She’s right.

Yeah, you can't draw back the curtain for normal people to read what goes on here 😉

 

[deezid]

On 7/16/2022 at 7:01 PM, Jay60p said:

In that case, perhaps a custom LUT could adjust the Fuji dark reds to a darker richer red,
closer to the Canon look.
Beyond my abilities unfortunately.

Impossible to recreate that hasn't been there in the first place with just a lut, some good tracking and color grading skills are necessary to create the illusion of chroma detail/tonality on Fuji cameras.

Sad to see the issue still hasn't been fixed since the X-T3. I also think it's chroma noise reduction related as well, looks like there are tons of spatial chroma filtering applied.

[PannySVHS]

"Gut and schoen" @deezid Now you are needed in the GH6 thread to answer all our questions. 🙂 Our friend @hyalinejimhas been very busy already to do so. cheers🙂

[Jay60p]

On 7/18/2022 at 11:02 AM, Attila Bakos said:

@Jay60p This issue is mostly about video, and at this point no further tests are needed, the color loss is there, and the reason is known.

Got it! It’s not color science, not NR, it’s compression. (The missing red leaves)

My apologies if you posted about compression already.  
As my brother would say “Just hit me in the head!”

Differing Compression levels explain both my indoor low complexity/high detail Cr shots,
and your outdoor high complexity/low detail Cr very neatly, as well as my high complexity foliage shot yesterday - it looks like yours. (My earlier indoor tests were during winter time).

Basically, when large amounts of detail must be discarded Fuji decided to smooth the reds more than Canon.
Maybe this allows for more green or luminance detail for any chosen data rate.

NOTE:
For people reading forums who are shopping for a family camera,
all the new cameras are great.
You can safely ignore everything we say. We are all crazy.

[FHDcrew]

On 7/18/2022 at 2:43 PM, Jay60p said:

Postscript:
My wife tried reading my last post.
She didn’t get very far.
She said we’re all crazy.
No doubt, She’s right.

Haha!  Isn’t that the reality we face 😆 

[Attila Bakos]

11 hours ago, Jay60p said:

Maybe this allows for more green or luminance detail for any chosen data rate.

In my video there is a section where I create a 100Mbps H.265 file from a 3584 x 1730 5D3 raw source and compare it with the 400Mbps Fujifilm file. While it's slightly less pixels than Fuji's UHD, the 100Mbps one is close to indistinguishable from the raw and the Cr channel is very detailed. And I didn't use some well optimized x265 encoder, I created the 100Mbps file in Resolve with GPU acceleration. When you put this next to Fuji's recording which has 4x the bitrate it becomes clear that there must be some purpose to the chroma smoothing other than preserving bandwidth for the luma channel. And I also mention in the video that I checked many files from other manufacturers (some of them have the same amount of luma detail and less bitrate) and didn't see this kind of chroma processing.

[Jay60p]

6 hours ago, Attila Bakos said:

In my video there is a section where I create a 100Mbps H.265 file from a 3584 x 1730 5D3 raw source and compare it with the 400Mbps Fujifilm file. While it's slightly less pixels than Fuji's UHD, the 100Mbps one is close to indistinguishable from the raw and the Cr channel is very detailed. And I didn't use some well optimized x265 encoder, I created the 100Mbps file in Resolve with GPU acceleration. When you put this next to Fuji's recording which has 4x the bitrate it becomes clear that there must be some purpose to the chroma smoothing other than preserving bandwidth for the luma channel. And I also mention in the video that I checked many files from other manufacturers (some of them have the same amount of luma detail and less bitrate) and didn't see this kind of chroma processing.

Thanks for that detailed response!
I believe every word you said,
but I still don’t know what you meant by “the reason is known”.
Have you come to a conclusion?