Jump to content

Attila Bakos

  • Posts

  • Joined

  • Last visited

Posts posted by Attila Bakos

  1. To anyone who is interested in recording ProRes RAW with the X-H2s, I just got confirmation from Fuji that they convert X-Trans to Bayer in the camera, as I suspected. Because the two patterns are different, you have to calculate what's not there. So let's say at a given pixel X-Trans stores red but for Bayer we need blue. I asked Fuji if they interpolate this blue using the surrounding blue values in the X-Trans pattern or they just copy the value of a blue neighboring pixel. The latter is faster but more prone to artifacts. Unfortunately I received an one-liner that this is proprietary info.
    I'll know more when I test this out myself, I'll receive a Ninja V+ in the following days.

  2. 6 hours ago, PannySVHS said:

    Would love to hear your opinion about the GH6! @Attila Bakos

    External recording with the Fuji is fine though? Thanks for your findings and contributions!

    I don't have the GH6 unfortunately, but I checked a ProRes HQ V-Log L sample and the chroma channels are very good.

    The Fuji chroma smoothing is there in all internal and external recordings, the only exception is ProRes RAW. As I mentioned earlier it's weird though. ProRes RAW can not hold X-Trans data, it's designed for Bayer. Atomos already confirmed to me that they don't modify the raw feed. I'm still waiting for Fuji's answer, but the only thing I can imagine is they demosaic X-Trans in the body, then remosaic it into Bayer and send that to the recorder. While it works it is kind of a trick, and when we compare ProRes RAW features to BRAW, one of the main things that sets them apart is that ProRes RAW is not debayered and thus more raw-like, while BRAW is partially debayered. So I'm not sure if we can talk about being not debayered as an advantage here, as the raw feed that the Ninja V records is actually already processed by the camera.
    And if you go into details, demosaicing is basically filling the "holes" in the R,G,B planes by interpolation methods, using the pixels that are known. See the "holes" here (it's for Bayer, but you'll get the point):

    Now if you remosaic the R,G,B planes to Bayer, you have to throw away about half the originally known values, as the pattern is different. You exchange them to interpolated values and send them to the Ninja as if they were the known values from the sensor. I have my doubts about the quality of this method, but I will test it when my Ninja V Plus arrives.

  3. I read some comments about the 1080p quality being worse on the X-H2s than on the X-T3/4, and I thought it's nonsense, but after a few quick tests it is worse indeed, for some reason the moiré/aliasing artifacts are more pronounced. Not sure about the reason yet, maybe the readout is different.

  4. Recently purchased the X-H2s, and you know me, the first thing I look at is chroma resolution. Just dropping this in, on the left side you see the Cr channel of an EOS R 4K C-Log 8bit 4:2:0 recording (about 420 Mbps), on the right side it's X-H2s 4K F-Log 10bit 4:2:2 (about 615Mbps):


    Don't get me wrong it's better than the X-T3, but this smoothing can still lead to loss of color. I really wish we could turn it off and deal with color noise in post.

    ProRes RAW helps but it's a mystery to me. RAW on the X-H2s has the X-Trans pattern, obviously. However, it gets magically converted into Bayer in the ProRes RAW file. I had discussions with people who are involved in ProRes RAW->DNG conversion, also did my own test with Raw convertor (resulting DNG has Bayer pattern) and it seems that the data that arrives to a Ninja V recorder is already Bayer, even though the sensor is X-Trans. I'm still trying to confirm all this but if the body demosaics the X-Trans pattern and remosaics it into Bayer, then it's kind of a hack which in some cases can lead to artifacts.

  5. 10 hours ago, Daniel Robben said:

    Thanks, yes, I reckonized but thought that's normal for non-all-intra frame encoding like long GOP. From my head I believe in prores first frames were same quality as following.

    Ah yes you are probably right 🙂

  6. On 11/22/2022 at 10:08 PM, Daniel Robben said:

    I checked this chroma smoothing issue on my X-H2s and I noticed that the chroma channel details are far better preserved when NOT using f-log or f-log2.

    Luma channel seems to perform the opposite, however not as significant.

    Yes the log profiles are worse. Btw when you're testing chroma components in Fuji files, make sure to skip the first few frames, those are worse than the rest.

  7. On 11/21/2022 at 8:51 PM, PannySVHS said:

     @Attila Bakos has recognized the Fuji XH2s having a lack in colour resolution, blurring the red channel.

    It's not the red channel that's blurred but the chroma components in the YCbCr format, that is Cb, and Cr. Both Cb and Cr are blurred the same way, but I used Cr for demonstration as it shows the issue a bit better. Btw Cr is red minus luma, so you are not that far from the truth 🙂

  8. 4 hours ago, herein2020 said:

    I believe you are referring to Intel Arc, that's the only card I am aware of that can do that. The problem is if you already have a desktop that is not QS capable then the only way to get a QS capable system is to buy or build a new one.

    No, I'm referring to NVIDIA/AMD. I just did a test for you. I have a Lenovo Legion 5i Pro laptop with an i7-12700h CPU and a 3070Ti GPU. I loaded a 2160p 4:2:2 HEVC F-Log clip to Resolve, added a LUT and some tweaking.
    With hybrid mode enabled (that is Resolve can use both internal GPU and the dedicated one), I get smooth 25fps playback with low CPU usage and rendered out the clip with the H.265 Master preset in 18 seconds.
    Now if I switch to dedicated GPU only mode, I get 21-23 fps playback and high CPU usage and the clip rendered out in 41 seconds.
    So basically with 10th or higher gen Intel CPU I get 4:2:2 support in addition to what the GPU already offers, no need for a Mac (for this).

  9. 4 hours ago, herein2020 said:

    Yes but to get that you need to buy a whole new PC, not an option to me and if you study Puget Sound's results, QS is still only marginal; meaning if you add DR color grades, and throw in a little Fusion and you are back to where you started; definitely not M1 Mac competition. There's just no way I am buying a whole new PC when all I need is a video card. 

    But what if you have an Intel processor with 4:2:2 decoding support and a dedicated GPU? I was under the impression that they can work together.

  10. 20 minutes ago, herein2020 said:

    It is very disappointing that in 2022 PC users still have no video card options for 4:2:2.

    That's true, however Intel QuickSync supports hw encoding/decoding of 4:2:2 HEVC since Gen10. Better than nothing 🙂

  11. 1 hour ago, Dustin Jenkins said:

    In Gimp there is a Colors menu and in that is Component>Extract Component and in there you can extract many different channels, one of which is "YCbCr Cr" and it extracts the Cr channel from YCbCr (I assume, at least. Seems pretty self explanatory)

    How do you import the video into GIMP? We can move to private message.

  12. 8 hours ago, Dustin Jenkins said:

    Are you selling your ycbcr extractor dctl? I'm using GIMP but it doesn't have a contrast and pivot and I don't feel that I'm getting the same results that you are.

    No I'm not selling it yet, but I validated my results with ffmpeg as it can extract the chroma channels as grayscale and I can safely say that my results are correct.
    How do you extract the channels in gimp?

  13. Here is a short demonstration of the Cr channels: https://drive.google.com/file/d/1veiqGodXKxerdNF_ksqLlAnFWbGQUtEe/view?usp=sharing

    8bit is rubbish, never ever use it. 10bit is somewhat better than on the X-T3/4, but the chroma smoothing is still there, ProRes doesn't seem to help either. Interestingly the X-H2s seems to have more smoothing than the X-H2.
    As much as I love Fuji for stills, I think I'll stick to other brands for video as I'm very sensitive for this kind of stuff, but for 99.9% of Fuji video shooters this is still a non-issue I guess.

  14. 2 hours ago, Dustin Jenkins said:

    Sure. I don't have 2 copies of the same lens so I won't be able to shoot them simultaneously.

    I might be able to get some black magic shots too, I don't own any canon or Sony since I've gone to Fuji.

    It's not important to shoot at the same time for this test, if the light is more or less the same then it's fine. When you shoot with other cameras, please use a format that's encoded in YCbCr (h.264, h.265, ProRes for example), as raw won't have this problem.

  15. 23 hours ago, Dustin Jenkins said:

    I have the x-h2 and would be happy to shoot anything in any resolution. I'm shooting mostly h.265 at 720mb but will gladly post anything you want.

    Please shoot something with lots of fine detail and varying colors at infinity focus (trees with autumn colors are the best, check my youtube video I posted in this topic), both h.265 and ProRes. When it comes to h.265, the higher bitrate the better, and ProRes should be 422. UHD or higher resolution is fine.
    When comparing both cameras, please use the same scene with the same settings. F-Log and/or F-Log2 please.
    I know I ask a lot, but if you have other cameras as well, Canon/Sony/Panasonic, it would be nice to see them as well. The issue might not be obvious until you start to compare it with other cameras.

  16. 1 hour ago, Andrew Reid said:

    Here is a ProRes LT clip at infinity


    Will try and get more footage today but the weather has been diabolical.

    Finally some noise in the Cr: https://www.dropbox.com/s/mesy1evzib28hb2/DSCF0234_Cr.mov?dl=1
    Noise is good in this case.

    When you have a chance to do more, please use 422 instead of LT, I was told that the difference in quality was noticeable in the X-H2(s), and minimal between 422 and HQ.

  17. I downloaded the 8K video sample from the bottom of this article: https://www.photographyblog.com/previews/fujifilm_x_h2_photos
    It's 7680x4320 10bit 4:2:0 H.265 29.97fps, but only 100Mbps.
    4:2:0 means that the chroma resolution should be 3840x2160.
    So I put the clip on a 3840x2160 timeline, and on the right side you see the Cr channel with added contrast for better visibility: https://www.dropbox.com/s/e04fm6oc5ac9qbz/X-H2_chroma.mov?dl=1

    I'll let you judge this, but I don't like what I see. Higher bitrate might help, or ProRes as well, but the chroma processing is still here with us 🙂 And this is not even F-Log where the issue was more prominent in the past.

  18. 16 hours ago, Jay60p said:

    but I still don’t know what you meant by “the reason is known”.
    Have you come to a conclusion?

    I meant the reason for the color loss, which is the chroma smoothing. The reasons for the chroma smoothing at this extent I don't know and as you said, we may never know.

    14 hours ago, Jay60p said:

    What I means is, you also see it as a compression issue?

    I wouldn't say this is a compression issue. This is not something caused by compression algorithms, this looks like some kind of smoothing, tupp already shared some thoughts about that in this thread:


  19. 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.

  20. @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.

  • Create New...