Arch Linux

bobadigo

Member

Registered: 2016-09-17

Posts: 34

[SOLVED] Color Banding

Good morning.
I have a little problem. Some images are affected of color banding. During the installation i selected "xf86-input-libinput" instead of "xf86-input-evdev" because i read that it is better in pretty everything. I also selected "libx264" and not "libx264-10bit", just because i have an AMVA+ panel (GW2470h) and it doesn't support 10 bit colors (I don't know if it do that with dithering). And if I am not wrong, libx264 is only for h264 encoding (and decoding, and not for the normal desktop usage).
So, why is there this problem, if i am running a 16.8 millions color monitor, that is the same color depth as full RGB scale (0 ~ 255 range for three colors = 2^24)?

EDIT

The same image, as the default desktop viewing have the problem but.. Now that I have tryed to view the image from telephone (Galaxy S5, that have a SuperAmoled display), it doesn't seems to be same: this is like without any band!
So... What could be the problem? The HDMI cable? Or what?

Last edited by bobadigo (2016-10-10 14:02:40)

Awebb

Member

Registered: 2010-05-06

Posts: 6,286

Re: [SOLVED] Color Banding

Please explain, what you mean by "color banding". Are you talking about that oval area? Try a different desktop environment (or plain Openbox). If the problem goes away, then whatever desktop you are using needs a setting.

Yes, libx264 and libinput have little to do with the way your desktop is being drawn.

ewaller

Administrator

From: Pasadena, CA

Registered: 2009-07-13

Posts: 19,796

Re: [SOLVED] Color Banding

So, why is there this problem, if i am running a 16.8 millions color monitor, that is the same color depth as full RGB scale (0 ~ 255 range for three colors = 2^24)?
http://postimg.org/image/letj514yt/

Yes on all counts.  But, as you are displaying (mostly) gray, you only get a 256 unique levels.  Yes, your picture is a desaturated color image, but for a moment pretend it is grey.  For grey, Red==Green==Blue.  For an image that is 1024 pixels wide, you would expect to see a step in brightness every 4 pixels.  One could cheat and change the red level in one column, then the green in the next column, and blue in the next.   The eye would do a pretty good job of integrating the brightness and probably would never notice the phase shift in the hue.    I doubt that the software you are running is doing that, so you are pretty much going to get 256 steps when approximating any flat field (grey (r=g=b), red, blue, green, cyan(b=g,r=0), yellow r=g,b=0)

Last edited by ewaller (2016-09-19 21:36:48)

---

Nothing is too wonderful to be true, if it be consistent with the laws of nature -- Michael Faraday
Sometimes it is the people no one can imagine anything of who do the things no one can imagine. -- Alan Turing
---
How to Ask Questions the Smart Way

Awebb

Member

Registered: 2010-05-06

Posts: 6,286

Re: [SOLVED] Color Banding

Is this only during the phase out effect for things like the shutdown menu, or is it always?

bobadigo

Member

Registered: 2016-09-17

Posts: 34

Re: [SOLVED] Color Banding

Please explain, what you mean by "color banding". Are you talking about that oval area?

I don't know how you call it.. I searched on wikipedia and i fount that: https://en.m.wikipedia.org/wiki/Colour_banding

Try a different desktop environment (or plain Openbox). If the problem goes away, then whatever desktop you are using needs a setting.

I also tryed KDE, and have the exactly problem of GNOME.

Yes, libx264 and libinput have little to do with the way your desktop is being drawn.

What do you mean with that? I didn't really get

So, why is there this problem, if i am running a 16.8 millions color monitor, that is the same color depth as full RGB scale (0 ~ 255 range for three colors = 2^24)?
http://postimg.org/image/letj514yt/

Yes on all counts.  But, as you are displaying (mostly) gray, you only get a 256 unique levels.  Yes, your picture is a desaturated color image, but for a moment pretend it is grey.  For grey, Red==Green==Blue.  For an image that is 1024 pixels wide, you would expect to see a step in brightness every 4 pixels.  One could cheat and change the red level in one column, then the green in the next column, and blue in the next.   The eye would do a pretty good job of integrating the brightness and probably would never notice the phase shift in the hue.    I doubt that the software you are running is doing that, so you are pretty much going to get 256 steps when approximating any flat field (grey (r=g=b), red, blue, green, cyan(b=g,r=0), yellow r=g,b=0)

I understand all, but what can be the problem so? If i see the same picture without that problem on a telephone, it mean that the problem is with the display configuration. But.. How can I effectively fix this issue?

I added a photo in the first post, at the end

EDIT

Is this only during the phase out effect for things like the shutdown menu, or is it always?

I noticed only in that "menu" but, it is (at least I think) always, considering the difference between two devices.

EDIT 2

If it can be useful, I'm using Intel integrated graphics (precisely the HD 530), with "xf86-video-intel".

Last edited by bobadigo (2016-09-19 22:00:48)

bobadigo

Member

Registered: 2016-09-17

Posts: 34

Re: [SOLVED] Color Banding

I tryed changing color profile in "Settings->Color" on Gnome.. Always the same.. What can I do?

I also checked out the Display settings and, on refresh rate there are two 60 hz option: 60 Hz and 60 Hz (NTSC). What's the difference?

I also noticed a awful screen tearing scrolling firefox pages.. Is that normal?

ewaller

Administrator

From: Pasadena, CA

Registered: 2009-07-13

Posts: 19,796

Re: [SOLVED] Color Banding

If i see the same picture without that problem on a telephone, it mean that the problem is with the display configuration. But.. How can I effectively fix this issue?

Does the phone have a greater color depth?  Maybe the phone display is small enough that your eyes do not resolve the color steps.  Look at it with some magnification.
Maybe the phone has a lower resolution.  Anyway, the brightness steps are closer together -- but they are there.

A 24 bit panel (8 bits/color) will give you 256 levels of brightness for a flat field.  As I said, you can play some games with dithering the LSBs at the cost of minor color error.
Sorry

---

Nothing is too wonderful to be true, if it be consistent with the laws of nature -- Michael Faraday
Sometimes it is the people no one can imagine anything of who do the things no one can imagine. -- Alan Turing
---
How to Ask Questions the Smart Way

bobadigo

Member

Registered: 2016-09-17

Posts: 34

Re: [SOLVED] Color Banding

The screen resolution is FHD, as the monitor, and the color gamma should be the same.. Probably is how you said: the lower screen size.
So, how can i do this trick? I don't really know how.
Anyway, I'am still not convinced that is a software problem, and not the size. I say that because the color band aren't perfectly scaled, but like: grey, dark grey and again (a bit brighter) grey, dark grey..

P.S. Is "libx264-10bit" ONLY for h264 enconding/deconding, as I said in the first post?

Last edited by bobadigo (2016-09-20 17:36:48)

Awebb

Member

Registered: 2010-05-06

Posts: 6,286

Re: [SOLVED] Color Banding

P.S. Is "libx264-10bit" ONLY for h264 enconding/deconding, as I said in the first post?

Look at the list of packages depending on it: https://www.archlinux.org/packages/extr … 4/libx264/. Doesn't look very desktop-y to me.

bobadigo

Member

Registered: 2016-09-17

Posts: 34

Re: [SOLVED] Color Banding

I didn't find the problem yet. I mark this topic as solved, hoping that it is a GPU problem (at the moment I'm using the integrated one.. I will buy a dedicated soon).
Anyway, thanks for all answers.

Last edited by bobadigo (2016-09-22 11:33:56)