Adventures in the PCU

Samuel H

Samuel H

Major Contributor
This is a thread for everything picture-control related
A twin of this other thread over at the Canon subforum: http://www.dvxuser.com/V6/showthread.php?262094-Adventures-in-the-PSE


And I'll inaugurate it by solemnly presenting the beta-2 of my Flaat suite of picture controls for Nikon DSLRs:
http://www.similaar.com/foto/flaat-picture-controls/index.html

Forget about the ultra-high dynamic range endeavours: this is now mainly about the everyday tools, Flaat_10 and Flaat_11

As with the Canon picture styles, my aim when creating these was:
* Low noise in the final images, achieved by efficiently using the codec's color space
* Easy to grade, achieved by generating a smooth, log-like response to light
* Leading to nicer skin tones, achieved by using Portrait as base picture control
* Only on Flaat_11/12, more recorded DR (less part of the sensor's DR being clipped in the shadows; this comes at cost of noise)


Would you give them a try and tell me what you think?
(for those who don't know, the instructions on how to install this are in the download page, following the link above)
 
Last edited:
could you in your wisdom make a profile that is catered to detail, low noise and good color representation for the shooter that for some jobs they dont want to spend time grading? For this kind of shooting I am shooting at iso 100-200 but shooting landscapes.

Cheers,

Jon
 
I'm not really into the "no grading" thing... if you can live with "easy to grade", you can go for Flaat_10p, but even then it's not perfect for what you want: in this version, everything is based on Portrait, because that's the path to nicer skin tones; for landscapes, my guess is that this will have color issues that you won't like

when I get some feedback on beta-2, I'll start working on the next version; if all goes well, that should include two suites, one based on Portrait and one based on Neutral; give it a month or so
 
Hi Samuel: Interesting stuff! Thanks for researching this.

I have a question (from the Dept. of Stupid Questions):

Might there be a chance that in-camera settings similar to what you're working on may also benefit external video recording via the D800's HDMI connector? Even theoretically? Just curious. Cheers!
 
I have no idea: I HAVE footage recorded externally with the D800 and Flaat_13p, but my (windows) editing software doesn't want to play with it :(
I can upload it to the server if anybody is interested

my guess is that:
* the day-to-day options (Flaat_10 and Flaat_12) will be great for external recording, as they aim for nicer, easier to grade images, not crazy transformations
* the crazy-DR options (specially Flaat_13) will help more with in-camera recirding than with external recording: the advantage of applying this aggressive curve in-camera instead of in-post is that it is applied to uncompressed 422 images, before the codec reduces them to 420 H.264; if you're going to save high-quality 422 anyway, there may not be a lot of difference between applying the crazy curves in-camera or in-post


edit:
the first Flaat_13p beta-2 test is up: http://vimeo.com/39899742
it's still about the crazy-DR, I hope I'll soon have some footage with the regular variants too
 
Last edited:
Hi Samuel H,

Thanks for all the detailed research and providing us users with some great picture styles.

Steve.
 
new test footage, with Flaat_10, Flaat_11, Flaat_12: http://vimeo.com/39937102
it's still mostly a DR test, but I found it very useful: I'm very glad to see Flaat_12 seems definitely useable, I wouldn't say it's any more noisy than Flaat_11 on the Canons
 
Thanks again for your work on the flaat picture profile for the D800. Don't you think that it would have been better to shoot at base ISO (100 ISO) to see the best less noisy shadows you can have. The second thing if there is a 1080p version or is it a 100% crop of a 1080p image just to see the noise because it seems very low at this rez.
 
hey Samuel H
thank you for you hard work and research on flaat picture profile , i was wondering if you did also test on the D4 dr and picture profile , and if so do you find it any different than the d800 ?
 
@Danielvilliers: I didn't run the tests; the last Nikon I used was shooting film, 15 years ago; I relied on users to give me the files that I needed to create the profiles (using only synthetic tools), and now I rely on them to run the tests too (that's the main reason for this to be a beta version) - so far everything I've seen is 720p; I would love to see some skin tone tests, and DR tests at ISO 100 and 1080p :)

@t.p.: I have sent the profiles to a couple of D7000 and D4 users, but I haven't heard from them yet; they should work, but they have been custom-tailored for the D800; in order to confirm that they don't need to be modified to work properly on the D4/D7000, I would need to get a very specific (but easy to shoot) set of RAW stills; send me an email (feedback.similaar at gmail.com) if you're interested and have some time for this
 
With my experience with the hacked gh2, the higher bitrate produced very fine an tight film grain type of noise. They are much less obtrusive that the standard 24 mbit codec does. I am think that the higher bitrate from the clean hdmi could produce the same thing and allow something like flaat_12 to be much better for the noise in the shadows.
 
for flaat_12, that might work

for flaat_13, I think it wouldn't: the problem with that one is that the camera applies the curve defined in the picture control over an uncompressed 8-bit 422 image; that image has very little information in the darkest shadows, and the curve is very aggressive, and this is what creates the heavy noise you see in this example: http://vimeo.com/39899742

it's a problem that arises on an uncompressed 8bit 422 image, before the final codec does its work; no increase in the codec is going to improve that issue

what could help here is a change in the debayering parameters, and it's not easy to affect those

I've run some tests on the old Canons using a couple special Magic Lantern versions that a1ex sent me, with menus to change some of these parameters, but we still don't know if they work as they should, or we're affecting something else; we'll have to wait until the 5D3 gets the ML treatment (in the old Canons, the sensor's DR coincides with the debayering bounds; the 5D3 is the first Canon video DSLR that has significantly more DR than what a "simple" picture style can get back)
 
Was it filmed at 720p because I think this mode is not very good in the D800. In fact when I look at the video below even with no special picture profile, this camera has some damn good dynamic range already. Hopefully we will see some example of Flaat in normal shooting condition and will be experiment when I get the camera.


 
yes, everything I've seen so far is 720p; for the Flaat_13p test, I know it was shot that way (because of issues with the external recorde, although those issues in the end made the external recorder footage useless, and the video you can see is all internal-footage); the other ones, maybe shot at 720p, maybe shot at 1080p and then downscaled, no idea
 
I think I read that the hdmi signal turns to 720p if there is a card in the SD or CF slot. That is you can't record simultaneously in camera and clean 1080p using hdmi. I hope the tester knows this. So there are still hope that we might get better result from the 1080p output or clean hdmi one.
 
Ok, so I run all my tests on the D7000, and I can say: Flaat works on the D7000 too, but it's definitely not ideal, light response is very different (highlight rolloff is quicker, the kink into the shadows becomes a bit too sudden, and my DR ratings no longer work, even if the sensor's DR is not really that different).

So, you can use them (and they'll look good in the midtones, where light response is still near-log), but it's not ideal, I would have to create a custom-tailored version of Flaat for the D7000 (and I'm not sure that's all that interesting anymore).

And use only the ones with the highest DR, because it seems to get about 2 stops less DR than the D800 with each profile.

No idea about the D4 yet.


edit:
highlight rolloff, transition into the shadows, contrast in the midtones... everything is different!
and should I say: in every case, I prefer what the D800 does
great improvement in the in-camera treatment of the image
 
Last edited:
Version 1 of Flaat for Nikon is out!
This is the first non-beta version, I've had a D5200 with me for a couple of weeks, so this time I've been able to test it thoroughly.

What to expect:
* Same colors (based on Portrait, so: better skin tones than with Neutral-based profiles)
* Less noise in the shadows (I simplified the curve and raised the black point, I think this cleans up the shadows a little bit, please tell me what you think)
* Similar DR (the camera won't give me any more), maybe slightly better in the highlights (very slightly), and also in the shadows if you think in terms of "usable"

Tests here: https://vimeo.com/67542641
And more here: http://www.similaar.com/foto/flaat-picture-controls/index.html
 
Thank you for your work. Do you think that version 1 of Flaat will work on the D7100? I have just got a Pix 220 and I'm going to do some filming at the end of the week, I'd love to try out the Flaat for Nikon.
 
Yes, it will work perfectly fine on the D7100. No idea about the 220, does that even take custom styles? (if not, I would go for portrait with minimum contrast).
 
You must log in or register to reply here.
Back
Top