Comments on: (Tutorial) Gamma Correction: A Story of Linearity

By: sfsdf

sfsdf — Sat, 09 Mar 2013 14:34:28 +0000

We need a way to know which gamma a monitor expects to do this right?
Sounds like a job for a HDMI standard and others.

By: Lightness1024

Lightness1024 — Mon, 16 Apr 2012 08:47:10 +0000

The last two comments: seconded.

By: waldyk

waldyk — Tue, 19 Oct 2010 23:29:53 +0000

Well, I’ve always been increasing gamma values for my CTX CRT monitor, because otherwise it was not only too dark (i remember that pain when playing shooter\action game, the dark scene – and u don’t see anything and then u get killed), but also realism suffered as well – trees, landscapes appeared artificially saturated, skin tone appeared sometomes too red, sometimes yellowish, overally the image looked like underexposed photo. Mostly affected by this issue are whose Trinitron tubes.
Now, owing a notebook with lcd display, I must say what u absolutely don’t need any upping of gamma levels, moreover i actually lowered my gamma setts for it, because otherwise it’s not just too bright, but, given the fact it’s TN, colors in the bright areas are literally saying undistinguishable! So if u got LCD with TN matrix in it, u should try lowering gamma – it will visually improve picture AND if u own a hi quality PVA/MVA or IPS one then u should no worry at all – in most cases gamma and colors in them are precise enough. Also i must tell that none of the monitor types (LCD vs CRT, TN vs IPS vs *VA) do any gamma correction – all differences are just because of differencies in technologies of picture formation on the screen.
P.S. Sorry for mistakes/misspelling (East Europe! :D)

By: Maurice

Maurice — Mon, 04 Oct 2010 06:54:38 +0000

In these examples you are computing a linear image, while your display device expects a non-linear image. That’s why the image look too bright. Gamma correction is used such that a pixel intensity of 128 looks approximately twice as bright as 64. The latter is one of the assumptions in the Phong shading model you use in your examples, yet it no longer holds in the images you display.

As I understood things, gamma correction often refers to performing arithmetic with colors (like averaging, or adding as you mentioned) in linear space. That means you transform the input colors to linear space (^2.2), average them, and then convert the output back to non-linear space (^(1.0/2.2)). That’s because with averaging you should indeed take the actual amount of light emitted by the display device into account, not the perceived brightness.

By: JeGX

JeGX — Sat, 02 Oct 2010 14:09:35 +0000

@Dave @Stefan: I fixed the bug in the GLSL shader of Gamma_Correction_Demo.xml. You can download again the demo. I posted some details about this bug here: TIPS: How to Quickly Test GLSL Shaders for ATI on a NVIDIA System?

By: [TIPS] How to Quickly Test GLSL Shaders for ATI on a NVIDIA System? - 3D Tech News, Pixel Hacking, Data Visualization and 3D Programming - Geeks3D.com

Sat, 02 Oct 2010 14:04:54 +0000

[…] usually test all my GLSL shaders on both architectures but for my last tutorial about gamma correction I was a bit lazy and I only tested on my dev system (GeForce GTX 460). Verdict: users with a Radeon […]

By: Stefan

Stefan — Sat, 02 Oct 2010 13:10:24 +0000

Same issue with ATI Mobility 5470.
Channel demo works fine though.

By: Dave

Dave — Sat, 02 Oct 2010 11:04:19 +0000

Gamma Correction Demo doesn’t work here (ATI mobility 5650). Errors in GammaCorrectionShader:
ERROR: 0:7: error(#202) No matching overloaded function found pow
ERROR: 0:7: error(#160) Cannot convert from ‘const float’ to ‘3-component vector of float’

Dave

By: Matumbo

Matumbo — Sat, 02 Oct 2010 09:21:24 +0000

I’m glad you talk about that 🙂

If you want your gamma corrected result to keep the same yellow color that appears in the non gamma corrected side, you can apply the pow(2.2) to your input color. That way you’ll see the exact same color, but with the sharp fallof and the beautiful spec highlight.

About DirectX10, I think it also applies the inverse gamma after the blending stage, so that the blending is also gamma correct.

By: FMoreira

FMoreira — Fri, 01 Oct 2010 22:58:40 +0000

XBOX 360 and D3D >= 10 hardware does this automatically 🙂
That’s why some times ago the first users of games that used D3D10 as the rendering API complained about brighter images 😛

By: jarik

jarik — Fri, 01 Oct 2010 21:38:00 +0000

I’ve played with vidoe-card gamma correction settings. Changing it from 1.0 to 2.2 caused image over-burned (too bright). Is it possible that LCD monitors do correction automaticly ?