Jump to content

Color detail issues in Fujifilm video files


Attila Bakos
 Share

Recommended Posts

EOSHD Pro Color 5 for Sony cameras EOSHD Z LOG for Nikon CamerasEOSHD C-LOG and Film Profiles for All Canon DSLRs

I don't know guys... just found two GFX100s F-Log clips here (200Mbps 10bit HEVC): https://drive.google.com/drive/folders/1UpKhgQpv8D6pEICjEBvLuAaSTSEvQaHX

And here is the result (download please): https://drive.google.com/file/d/1vjFxXTTaqftD8E8GN1vrT5ZTPkaUb6Nq/view?usp=sharing

The sequence is the following: original, Y, Cb (with added contrast), Cr (with added contrast)

Would be nice to see something more colorful, but I already don't know how this is any better from what we have seen from the X-T3 so far.

Link to comment
Share on other sites

1 hour ago, Attila Bakos said:

Another samples, this time from the GFX100, keep in mind that this is 8bit: https://drive.google.com/file/d/1fyCiDFhbkskhjxz2miFgLXn523uP80ns/view?usp=sharing

Sequence is the same as before: original, Y, Cb (with added contrast), Cr (with added contrast)

This might be the worst I've seen so far.

That (might) indicate it's not a bug, but a deliberate decision.

It all seems rather odd to me..  I get why Canon etc cripple their cameras, but why would Fuji do it?  and in such a strange way?  

My knowledge of X-Trans sensors is limited, but I thought they were simply an alternative layout of the RGB photosites in order to get an advantage when de-bayering.  Nothing in that seems to indicate they'd have some sort of horrendous chroma noise issue that would require such brutal chroma NR.

Link to comment
Share on other sites

  • Administrators
On 2/17/2022 at 6:14 PM, Attila Bakos said:

I found something strange when I was looking at the chromaticity channels of Fujifilm video files and did some search online, and to my surprise noone is talking about this, so I decided to make a video about it. If you're a Fuji user, you might be interested.

 

Great work. I wonder if the same issue is present on the GFX 100 in ProRes RAW to an external recorder?

Link to comment
Share on other sites

28 minutes ago, Andrew Reid said:

My theory is Fuji is doing this on purpose as a form of hidden noise reduction.

That's pretty much my conclusion too, based on the info in the Petapixel article Django linked to upthread.

Given the smaller percentage of red & blue photo sites versus green ones in X-trans (compared to a Bayer pattern), there is likely to be a higher level of chroma noise that needs to be dealt with - doing what looks like pretty simplistic spatial (and maybe some temporal) filtering is a 'cheap' way of suppressing it.

Link to comment
Share on other sites

  • Administrators

Good point about X-Trans. It's probably one of those pragmatic Fuji compromises like 4:2:0 in video mode. At least they are not chopping IBIS out of a $4500 EOS R5C! Could be worse.

Fuji's JPEGs and colour does have a nice smooth look most of the time... and now we know why? 🙂

It isn't necessarily horrible, but it clearly has downsides for certain kinds of shot when viewed at 1:1.

This is a great find though (and it's one of the best advertisements for Magic Lantern RAW I've ever seen).

Link to comment
Share on other sites

  • Administrators
On 2/18/2022 at 5:20 PM, Attila Bakos said:

I believe this has nothing to do with demosaicing, all the color information that's missing from the videos/jpegs are there in the x-trans raw files. I can show it tomorrow if necessary.

Edit: looking at Django's comment it's necessary 🙂

Ah so not X-Trans in fact.

I have noticed similar things with other cameras in foliage.

Remember the Samsung NX1 vs Canon 1D C video on EOSHD?

Are we sure it isn't a GPU decoder / NLE problem and it is definitely the camera?

Link to comment
Share on other sites

No, check out the Petapixel article I linked, this isn't a new finding and predates even video capabilities of the Fujis: it is known as the "Fuji watercolour effect". Fuji apply heavy chroma NR in their Jpeg/h264/h265 engine for their immune-to-moiré claims of the X-Trans CFA.

I assume they do the same in the GFX100 as that model does not have an optical low-pass filter despite being a Bayer sensor (Fuji claim its megapixel count is so high it can resolve complex textures with little to no moiré effects). Not sure if this chroma NR applies to GFX100 Jpegs or 4K video only.

Link to comment
Share on other sites

48 minutes ago, Andrew Reid said:

Are we sure it isn't a GPU decoder / NLE problem and it is definitely the camera?

I used ffmpeg first to extract the chroma channels into grayscale videos and I used that to validate the results of my plugin. They were identical so it's safe to say that it's not an NLE/GPU decoder issue.

Link to comment
Share on other sites

Again the root cause is indeed the X-Trans CFA. Fuji uses a demosaic algorithm for video/jpeg that prioritises speed and efficiency in a small camera body. This leads to colour smearing issue. It doesn't happen to X-Trans RAW stills anymore because modern RAW processing softwares have much more refined algorithms to deal with X-Trans.

Link to comment
Share on other sites

11 hours ago, androidlad said:

Again the root cause is indeed the X-Trans CFA. Fuji uses a demosaic algorithm for video/jpeg that prioritises speed and efficiency in a small camera body. This leads to colour smearing issue. It doesn't happen to X-Trans RAW stills anymore because modern RAW processing softwares have much more refined algorithms to deal with X-Trans.

I disagree on the modern RAW processing.
I'm at the moment switching from Adobe to Capture One.
It comes with a few extra pluses, like the ability to buy a perpetual license (a big plus in my book).

Link to comment
Share on other sites

A last sample, this is from the GFX100s, 4K, 400Mbps, 8bit H.264. It's not F-Log, possibly Provia, so it's chroma channels are supposed to be slightly better compared to F-Log: https://drive.google.com/file/d/1F4W_DYYkYlQIV49lakIGuz1L4xnB3oNm/view?usp=sharing

From top to bottom: Original, Cb (with added contrast), Cr (with added contrast)

I post this because there are some trees in the background.

The theory about this being an X-Trans only issue sounds fine, I might even believe it, but then all the clips that I found from GFX bodies so far tell a different story.

Link to comment
Share on other sites

6 hours ago, Attila Bakos said:

A last sample, this is from the GFX100s, 4K, 400Mbps, 8bit H.264. It's not F-Log, possibly Provia, so it's chroma channels are supposed to be slightly better compared to F-Log: https://drive.google.com/file/d/1F4W_DYYkYlQIV49lakIGuz1L4xnB3oNm/view?usp=sharing

From top to bottom: Original, Cb (with added contrast), Cr (with added contrast)

I post this because there are some trees in the background.

The theory about this being an X-Trans only issue sounds fine, I might even believe it, but then all the clips that I found from GFX bodies so far tell a different story.

Considering you have access to the camera, maybe you can do a noise test by taking a few RAW stills at a few ISO settings of a dark blank surface?

It would be interesting to see what the levels of noise are between the R, G and B channels.  If you have another camera from a different manufacturer handy then a comparison would be useful too.

Then we'll know if there even is a big noise problem with the sensor or not.  Maybe it's terrible and such heavy-handed processing might be understandable, even if the severity might not be justified.

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Restore formatting

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

 Share

×
×
  • Create New...