Bad demosaic - pink rules the white panel?
At please find an 100% crop of a screenshot out of Capture One; it depicts (in part) a large public display LCD / LED panel.
The panel's white is rendered all pink - that's "awesome" :/
I believe that the demosaic algorithm is ultimately responsible for this misrendering. The open source software "darktable", for instance, will misrender the raw file exactly the same way using the default PPG demosaic algorithm. Switching to the AMaZE demosaic algorithm will fix that panel from pink to proper white.
Is there any means within C1 to switch to a different demosaic strategy?
What else can I do within C1 to make the pink go away, with limited effort and good results?
And - what can I do in Real Life to not fall into such a trap? Not take photos of such panels? 😉
The panel's white is rendered all pink - that's "awesome" :/
I believe that the demosaic algorithm is ultimately responsible for this misrendering. The open source software "darktable", for instance, will misrender the raw file exactly the same way using the default PPG demosaic algorithm. Switching to the AMaZE demosaic algorithm will fix that panel from pink to proper white.
Is there any means within C1 to switch to a different demosaic strategy?
What else can I do within C1 to make the pink go away, with limited effort and good results?
And - what can I do in Real Life to not fall into such a trap? Not take photos of such panels? 😉
0
-
Have you tried the moiré adjustment?
Ian0 -
[quote="Ian3" wrote:
Have you tried the moiré adjustment?
Yes, unfortunately without positive results. I tried all conceivable combinations of Amount / Pattern.0 -
What was the shutter speed? 0 -
Shutter Speed : 1/640
This seems to be plenty fast; such displays would have a refresh rate / speed somewhere round 1/20s?0 -
Oh, and thinking of it - I had a (top quality) polarizing filter in front of the lens.
I'd be hesitant to point to the polarizer as the root cause, though, because using a different demosaic strategy (in other software - which I'd like to avoid) really "fixes" the problem.
Also, a polarizer would also probably introduce different artefacts - to show that pink (or magenta) colour, green would have to be *totally* filtered out, so the panel would have to emit green *exactly* on the right polarization plane and red/blue *exactly* off plane.
The small JPEG thumbnail embedded into the raw file (RW2) shows no false colours, by the way, just totally gross moiré 😄0 -
I also don't believe your polarizing is to blame. In my opinion this is purely an aliasing problem (detail greater than the sensor is capable of recording is incident on the sensor). Changing the Demosaic algorithm cannot eliminate aliasing (as far as I know nothing can once the data has been captured, hence the need for an AA filter) but it could mask some of the effects and make it less obvious. It would be nice if you could show us the same image but demosaiced using the other software.
I unfortunately do not know of any way to change the Demosaic algorithm in C1. If it was an image I wanted to save I would probably create a new adjustment layer and mask off the affected area. I would then use the colour editor to desaturate the offending colours. I would even consider applying negative structure and clarity to try and reduce the obviousness of the false patterns in the details. These suggestions obviously only really work in the case where the sign is basically B&W if it is displaying a colour image I think that would be quite the nightmare to fix.
A camera with a stronger AA filter would normally be more resistant (depending on strength even immune) to this type of problem.
Without changing camera you would effectively need to try and reduce the resolution of the detail landing on the sensor, this could be done in numerous ways:
1) Lens with less resolution
2) Standing further away (you need to be far enough that the lens limits the resolution)
3) Use a slower shutter speed hoping for some minor blurring (trick is to get the correct amount to not destroy entire image).
4) Use a filter that softens the image a little0 -
[quote="daffy" wrote:
And - what can I do in Real Life to not fall into such a trap? Not take photos of such panels? 😉[quote="daffy" wrote:
Shutter Speed : 1/640
This seems to be plenty fast; such displays would have a refresh rate / speed somewhere round 1/20s?
In my experience shooting projectors and such the higher the shutter speed the more likely refresh rate artifacts are to arise. If the refresh rate of the panel is say 1/60s and you're shooting at 1/120s then the camera only records half the cycle of the panel and artifacts appear. Move the shutter speed to 1/60s and the camera records the entire cycle of the panel and a more accurate image is recorded. Can't say for certain that these panels you are shooting behave the same.
Also I'd try it without the polarizer just for testing purposes. Strange things can happen with polarizers and shooting through plastics that are also polarized, ever try using a polarizer through an airplane window?0 -
[quote="gnwooding" wrote:
hence the need for an AA filter)
For the record, the Panasonic G9 which produced this has 20 mp on an m43 sensor (one quarter of the area of a full-frame sensor) - and it has no AA filter... The lens, well, "biting sharp".
Many thanks for your suggestions!
I will try to exercise the technique for the learning experience. Alas, to the left of that crop, a wonderfully touristy collage of Swiss mountains, a City of Zurich iconic building, and a Formula E race car is presented, in the most wonderfully light'ed colours imaginable 😉0 -
[quote="cdc" wrote:
Also I'd try it without the polarizer just for testing purposes.
Many thanks for the input!
I will try shooting this thing again when the next event comes up; it's summer, it can't be long 😊0
Post is closed for comments.
Comments
9 comments