abrupt highlights clipping

[Nick Williams]

I see, that makes sense. So we might possibly see better results coming out uncompressed. Thanks for the insight guys.

[Steven Thomas]

Actually,
The banding occurs during the time lapse footage, using the normal recording modes there's no issue. Well, other than the 8bit gradient capability itself.

Maybe someone will jump in to explain why this happening with the time lapse mode. This has been brought up before with image samples.

[Steven Thomas]

Originally Posted by Nick Williams

I see, that makes sense. So we might possibly see better results coming out uncompressed. Thanks for the insight guys.

Since the SDI is 10bit vs the in camera 8bit and Sony is utilizing all of the SDI 10bits, this would offer smoother gradients. But, since this particular banding is present in the time lapse and not normal record mode, I'm not sure how the time laspse will look via SDI.

[Nick Williams]

That's true, I figured that if an uncompressed signal may fix the banding issue that it probably wouldn't fix it on the time lapse side. I don't shoot a lot of sunrises or anything, so it shouldn't be a common problem for me.
But say that was the result of shooting normal speed.. would I adjust the knee or slope? To try and smooth those together? I'm just not familiar enough with those settings to know what to adjust. I guess if I wasn't shooting so close to blow out, that might help too.

[Piotr Wozniacki]

OK guys - back to the topic. I had a sunny day today, so - after my trials with trees againt gray clouds - I made some tests with beautiful, deep blue sky this time...

The 3grabs show the same scenery with:

- STD1, knee manual at 95
- CINE4 (unfortunately - still some "abrupt clipping", Adam !)
- CINE2 (this one IS safe, but really very, very flat - see the next post).

[Piotr Wozniacki]

And here is some more comparison:

- STD1 with auto knee, against
- safe CINE2

Obviously the auto knee seems to be doing better than manual at 95 (though this test is far from scientific, as the scenery is not identical - even more backlit; this made the zebra appear and allowed for iris adjustment while shooting). Therefore, below is again:

- STD1 knee manual at 95 (sorry the picture has wrong name - it actually IS manual at 95, not auto!)
- CINE2

Can you see the aureole around the roof and trees with STD1? And there was not even a trace of zebra there while shooting !!! Do you understand now why I think it's not good at all, and would do with some similars grabs of yours to compare?

The only gamma free of it is CINE2 -- but how flat it looks!

Go figure...

[Tom Roper]

Hi Piotr,

Just of few observations I would make, you may want to choke me for saying them, but they are my honest impressions, and a little more general, a little less specific.

I think std 1 auto knee 2 is clipped in both the luma and chroma channels. The exposure of this scene breaks the dynamic range. My judgment in this case is that cine 2 looks flat but only by direct comparison to cine1. On its own merit for video, (not a screen shot), I think it actually still remains on the margin for oversaturation in the chroma channels, and quite contrasty.

Std1 auto knee 2 is more excessive. Look at the blue cast in the shadows, and the color shift on the bricks. The gray and pink sunlit bricks turn to blue and purple.

I think it may be possible to increase the luma exposure in cine 2 to bring out some detail in the shadows, and not clip the highlights. Cine 2 shows less tendency to clip the chroma, thus the colors do not shift as radically when transitioning between sunlit and shadow areas in the midtones (i.e. the bricks).

I think there is excessive chroma saturation in all the screen shots. What that does is open the possibility for colors that bleed at the edges. I see some of that in your pictures. Your EX1, while being noted for lower CA than my XH-A1 for example, is nonetheless overemphasizing the CA that is there. Look at the edge along the roof gutter on the right in cine4.jpg, and along the gutter and rooflines at the upper left in std 1 auto knee 2.jpg.

One characteristic that distinguishes Canon's ugly red/green fringing from Sony is that Sony's fringing when it does appear, is purple/bluish/green. I think it is preferable usually but works against you when the backdrop is a blue sky. And some of the hue shift at the point of highlight clipping in the area of the branches may actually be minor CA that is overemphasized or clipped due to an overall oversaturation of the color channels. In other words, what begins as a minor optical blue fringing, when oversaturated in the chroma channel bleeds over into an amorphous bluish blob against a clipping sky highlight behind. I think I see minor blue CA that in some areas of the picture is clipping into prominence by overall high gain or saturation in the chroma channel.

...that's what it LOOKs like to me. You have the camera and vectorscope displays, so I may be all wrong, just sharing my visual observation.

Speaking on a personal preference, I very much like high contrast, color saturated images. And I have my XH-A1 presets set up for this generally. It may even be more "forgiving" in the sense, of color or exposure latitude. It does seem to more gracefully handle exposure/chroma gradients in the type of high contrast, bright outdoor scenes that you are testing. But that said...I believe the recent pictures above from you take contrast and saturation a bit too far. If it was me, I would try and go back to a baseline, decrease the saturation and dynamic range, and see if there is a decrease in the tendency for the colored highlights to bleed, and perhaps with that, the abrupt highlight clipping. I think you can still do that and still retain saturated high contrast video with punch. I think it can be a mistake to make too many judgments from side by side still images. As an anecdote, I used to sell hi-fi speakers as a junior, doing A-B listening comparisons. Invariably, people would purchase the "louder" sounding speaker even if the sound coming from it was colored. We can make the same mistake when doing A-B's on still images coming from a video camera. In my own work, (which is more play than work since it's hobby, not occupation), I constantly compare my video to video from others, Discovery HD Channel, HDNet, CBS etc. The eyes can fixate on something wrong, and overcorrect for it. Extended viewing of a broad range of material, visuals and sounds are what prevent fixations.

My judgment is that your colors are extremely vivid, to the point of imparting color casts in various parts of the image. Again, refer back to the bricks. It's not just a white balance issue. It's a case of shifting balance, bricks that show the correct color in the sun, taking on a hue in the shadows. I can't tell you what to do, only what I would do. And if it was me, I would try re-doing some of your tests at lower color saturation gains. Just my $0.02

Tom

[Tom Roper]

Just to add...


In the old days of film, we used to call color shifts caused by under or over expsoure "reciprocity failure." That's how I would describe std1 autoknee 2.jpg.

It works both ways, color shifts due to under or over exposure.

Louder isn't always better.

[Michael H. Stevens]

Piotr:

I think with all these pictures you are trying to sell your beautiful house!

Seriously, forget CINE2 unless for broardcast, you know it is just CINE1 truncated. For bright scene like you have shoot show us CINE1 or 3. I too was thinking why does everybody go HiSat. I have found on my scopes that the blue often clips when the luma is fine and HiSat makes this worse. Try C1 & 3 with normal or cinema matrix.

Mike

[Piotr Wozniacki]

Tom and Mike,

Thanks for your comments. You're absolutely right - these grabs were taken from my experimental shootings with PPs varying in the gamma used, but having all the same, highly saturated colour matrix. Even though I do like it contrasty, rich and juicy, some of them are indeed oversaturated even to my eye, and I have since modified the PP's accordingly.

However, with deep blue sky it was not that easy to clip the highlights as it was with dull wheather and overcast sky, so I made it more difficult for the camera by increasing the chroma... And the peculiarity of "abrupt clipping" appeared again - even with the Cine4 that Adam recommended as safe in this respect. Only Cine2 seems completely safe (at least as luma is concerned - chroma can still be clipping, as Adam pointed out).

Well, I guess I also owe some positive comments to this machine: were it not for abrupt clipping, the contrast range and colour saturation could be really considered absolutely fantastic. What needs to be said is that even with this level of colour gain (+25 with the Hisat matrix) as well as detail amount present, there is almost no mosquito noise or other artefacts that with the HDV codec on my V1E would have spoiled the image completely.

PS. Just received an e-mail from Adam with a couple of invaluable remarks and observations. One of the most important is that with Cine2 - even though it's safe in that its luma never exceeds 100% - the chroma is clipping nevertheless in the high 90-es. So, beware of that! This confirms your comments about the chroma (especially blue channel) clipping, which I can also see in Vegas scopes.

I will be trying now the Cinema matrix.

[Paul Kellett]

Piotr,what picture profile are you getting most satisfied with,i'm using bill ravens but i presume you've strayed a bit from that now.
Are you trying to get a profile which you can use all the time or different profiles for different situations,ie,profile for overcast day,profile for sunny day etc.
If you could post your profile(s) that'd be great.
Thanks,Paul.

[Piotr Wozniacki]

Paul, I am still experimenting to find my optimum PPs. When I'm ready, I'll post them.

No, I'm not using Bill's right now; all my colour matrix coefficients (pairs) are zeros (default). Perhaps I will fine-tune them later, when I settle down with the choice of matrices and gammas. Finally, I'll play with detail etc.

Yes, I always have a set of PPs for the most typical sceneries; here is the convention I've been using with cameras capable of fast switching between 6 PP's:

PP1: low-light indoors contrasty (STD1)
PP2: low-light indoors flat (STD4)
PP3: bright outdoors flat (Cine1)
PP4: bright outdoors contrasty/backlit (Cine4)
PP5: dull/low-light outdoors flat (TBD)
PP6: dull/low-light outdoors contrasty/backlit (TBD)

[Paul Kellett]

Thanks Piotr.
No Hisat on any of them,correct ?

Paul.

[Piotr Wozniacki]

Originally Posted by Paul Kellett

Thanks Piotr.
No Hisat on any of them,correct ?

Paul.

Actually Paul, as I mentioned earlier, the grabs posted today are ALL Hisat-based (plain at +25, with no individual pairs modification).

Whether or not my finall PPs will be Hisat I don't know yet - will try the Cine matrix over the weekedn (if the weather is like today :)

[Paul Kellett]

Thanks Piotr.