Luminous Landscape Forum
Equipment & Techniques => Digital Cameras & Shooting Techniques => Topic started by: Guillermo Luijk on May 26, 2018, 08:10:46 am
-
Whoever has taken pictures at a screen (TV, monitor, smartphone) probably has suffered terrible aliasing, moiré or even colour changes. The reason is we are sampling a discrete image with spatial frequencies in the same order of magnitude as our sensor:
(https://4.bp.blogspot.com/-82lfGiN5jig/WLhvfs2WvRI/AAAAAAAAA7I/f5-xRlCWk80_gwwjFuPqb-NuKtdl1LJQwCLcB/s1600/comparacion.png)
For a full HD TV set we have 1920*3=5760 LEDs in the horizontal direction, while a 24Mpx camera (6000x4000) has 6000 Bayer photosites in the same direction. Trouble is assured, specially if the sensor has a weak AA filter.
I have tested three atialiasing strategies on my A7 II comparing with the standard perfectly focused shot at f/5,6:
- Diffraction: using f/22 Fresnel circles as AA filter
- Camera shake: unorthodox and not very reproducible
- Defocus: using slight defocus as AA filter
100% crops:
(http://www.guillermoluijk.com/misc/tvcrop.png)
With this kind of artifacts (periodic patterns), it's important to check if they prevail once the image is reescaled. Photoshop is not very cautious regarding aliasing when downsizing, gaining some extra sharpness in exchange. 800px resizing:
(http://www.guillermoluijk.com/misc/tvoriginal.jpg)
(http://www.guillermoluijk.com/misc/tvdifractado.jpg)
(http://www.guillermoluijk.com/misc/tvtrepidado.jpg)
(http://www.guillermoluijk.com/misc/tvdesenfocado.jpg)
The first two images have important aliasing artifacts. Camera shake improves this quite a lot, but still displays banding in some areas so as the highest sharpness loss. Defocus seems to be the best option. Extra output sharpening would increase aliasing artifacts in the first images and would compensate for the defocused version drawbacks.
Experiences or thoughts?
Regards!
-
Two issues that cause aliasing (banding).
Initial capture. If it's aliased there is much less you do about it. Avoiding it can be done by defocusing or setting the aperture small enough that diffraction blurring kills it. You want just enough. My preference is aperture as it's much less sensitive than trying to get just the right defocus amount.
The second issue comes up when resizing. When color changes very rapidly or you have things like striped shirts, the low pass filters that are used prior to resampling can introduce aliasing (banding) where none had existed in the original image. Best way to deal with this is to convert to an RGB space with a gamma of 1. This must be done in 16 bits or you get other artifacts. So first I convert to 16 bits if necessary then convert to an altered ProPhoto RGB gamma=1 profile. Then resize, and convert back to a normal gamma colorspace.
-
Camera shake improves this quite a lot, but still displays banding in some areas so as the highest sharpness loss. Defocus seems to be the best option. Extra output sharpening would increase aliasing artifacts in the first images and would compensate for the defocused version drawbacks.
Experiences or thoughts?
Regards!
Aliasing and moire on lcd screens is as much trouble as rolling bars on CRTs used to be. And made more complex by the fact that since everything ends up on the Web, viewing screen resolution is uncontrollable.
Some thoughts:
- If the shot has people in it, it’s a lot easier to justify being focused on something other than the screen.
- Filters like Promist can do a lot to mitigate the sharpness that causes moire. You’ll get a bit of halation, but that often works well with a screen.
- Screen replacement is a PIA, but is sometimes a viable option, and you can precisely control the amount and type of blurring you use.
- If I need to shoot on-screen content, and screen capture is not a workable option, I try to shoot tight enough so that the screen pixels are large enough to mitigate the moire problem. It’s often a more impactful shot as well.
- Off-angle shots can be an interesting solution, though they can open another can of worms in that the scale is changing across the shot, and if sharp, moire can show up in some part of the image.
- For stills, focus bracketing is a good tool. (Nikon’s D850 would be a useful tool here.) For motion, not really an option.
-
I always photograph screens with my lens wide open. Or slightly defocused if the lens has the unfortunate property of being very sharp wide open. :)
-Dave-
-
Pixel shift helps a lot with the false color, but you have to watch out for temporal aliasing.
https://blog.kasson.com/a7riii/screen-captures-with-a7riii-pixel-shift/
Jim
-
Pixel shift helps a lot with the false color, but you have to watch out for temporal aliasing.
https://blog.kasson.com/a7riii/screen-captures-with-a7riii-pixel-shift/
Jim
I understand why pixel shift would mitigate false color, because it increases resolutionin all the color channels. But I don’t understand why going from 1/15th, which is already 4X the frequency of the monitor, to 1/5th has any effect. Can you elaborate on what’s going on there?
-
I understand why pixel shift would mitigate false color because it increases resolution in all the color channels. But I don’t understand why going from 1/15th, which is already 4X the frequency of the monitor, to 1/5th has any effect. Can you elaborate on what’s going on there?
The shutter speed was 1/25, not 1/15, which means that the exposure time is about half the scan time. Not sure exactly what's going on, but slowing down the shutter speed pretty much fixed the problem. It would be nice if someone with an a7RIII would try to reproduce my work. If they can't, maybe I have an uncontrolled variable. If that happens, I'll rerun the test with more controls.
Jim
-
Temporal dithering perhaps?
-
Temporal dithering perhaps?
That was my speculation, although I called it temporal aliasing in my post. But I don't really know for sure.
Jim
-
Whoever has taken pictures at a screen (TV, monitor, smartphone) probably has suffered terrible aliasing, moiré or even colour changes. The reason is we are sampling a discrete image with spatial frequencies in the same order of magnitude as our sensor:
(https://4.bp.blogspot.com/-82lfGiN5jig/WLhvfs2WvRI/AAAAAAAAA7I/f5-xRlCWk80_gwwjFuPqb-NuKtdl1LJQwCLcB/s1600/comparacion.png)
For a full HD TV set we have 1920*3=5760 LEDs in the horizontal direction, while a 24Mpx camera (6000x4000) has 6000 Bayer photosites in the same direction. Trouble is assured, specially if the sensor has a weak AA filter.
I have tested three atialiasing strategies on my A7 II comparing with the standard perfectly focused shot at f/5,6:
- Diffraction: using f/22 Fresnel circles as AA filter
- Camera shake: unorthodox and not very reproducible
- Defocus: using slight defocus as AA filter
100% crops:
(http://www.guillermoluijk.com/misc/tvcrop.png)
With this kind of artifacts (periodic patterns), it's important to check if they prevail once the image is reescaled. Photoshop is not very cautious regarding aliasing when downsizing, gaining some extra sharpness in exchange. 800px resizing:
(http://www.guillermoluijk.com/misc/tvoriginal.jpg)
(http://www.guillermoluijk.com/misc/tvdifractado.jpg)
(http://www.guillermoluijk.com/misc/tvtrepidado.jpg)
(http://www.guillermoluijk.com/misc/tvdesenfocado.jpg)
The first two images have important aliasing artifacts. Camera shake improves this quite a lot, but still displays banding in some areas so as the highest sharpness loss. Defocus seems to be the best option. Extra output sharpening would increase aliasing artifacts in the first images and would compensate for the defocused version drawbacks.
Experiences or thoughts?
Regards!
Yes, you're right