Display Adjustment

This page contains images I use to adjust TVs and computer monitors. I am not going to explain how to do this in depth, there are better references around. If possible, the first step is ensuring the display is not having a profile applied to it by software. There's many ways this can happen, too complex to go into here. Also set the monitor to have, if possible, a color temperature of 6500°K or sRGB. LCD displays must also be set to display at native resolution. With LCDs, viewing angle is very important for these adjustments, ensure your display is at your optimal angle before starting these adjustments.

White Level


This is adjusted by the Contrast or Picture control. Turn it up as high as possible, while still having the 254 checkerboard visible, if possible, and no color cast at all. At the very least, the 250 level should be visible. The background must appear as bright white.
white level test pattern


Black Level


This is adjusted by the Brightness or Black Level control. Turn down as low as possible, while still having the 2 checkerboard barely visible, if possible. At the very least, the 10 level should be visible. The background must appear as deep black.
Black Level Test

You may need to go back and forth adjusting the two levels until both are optimal.

Gamma


If your controls have a gamma setting, use it for the next adjustment. TVs usually cannot have their gamma adjusted. If your computer has no way of adjusting gamma, you could install QuickGamma to get the functionality. Move your browser window so the test image is centered in your display. The test image should look as close to grey as possible. If the image is coarsely striped, it is not being displayed at 100%, check your monitor and browser settings. Slight variations in color indicate a poorer quality graphics card. Try to get it as close to grey as possible.

This target looks totally grey when your display is using gamma 2.2, the industry standard, across the entire brightness spectrum. Any inaccuracies in your graphics card manifest as color casts.
Gamma 2.2 test
How does the gamma test image work? Answer

Summary


There's much more involved in getting your display to be completely accurate, but if you completed the above steps, your display is more accurate than 95% of the displays out there. As a final check, here is a reduced version of Digital Dog's test image.
Test photo image


How Gamma Test Image Works



gamma test closeup The test image works on the same principle as dithering and sub-pixel rendering, the human eye cannot discern fine detail, and instead averages adjacent detail elements. This is essentially how displays can show millions of colors using only 3 different colored LED elements. The blown up detail at left shows the colors composing the second vertical bar of the test image. You can see it is composed of dark magenta and light green stripes. Two adjacent stripes together have the same mean luminance as the adjacent grey to the left. The color levels are also calculated so they cancel each other out, resulting in an overall grey, but only at gamma 2.2. Any other gamma will result in one color dominating the other, giving the stripe a noticable color cast even if the gamma is only slightly off.

Actually, the test chart will display grey when your display is adjusted to the sRGB Standard. It is close to gamma 2.2, but it's calculation is more involved. The other complication is you cannot simply average pixel values to get the desired luminance. The pixel values have to be decoded to the actual luminance levels before any mathmatical operations can be performed, using the gamma-power function. Once the proper luminance level is determined, it is reencoded with the inverse power function to arrive at the nominal pixel value that we want. It is the application of the power function to one particular color pair of high and low values that causes a color cast to appear if the display gamma does not exactly match specification. The sRGB gamma-power function is:

L'=([(P/255)+0.055]/1.055)^2.4
where L' is the Luminance,
P is the actual pixel value

Example:
The grey on the left, near the middle, of the above detail is rgb(137,137,137).
The adjacent color stripes are rgb(137,187,137) and rgb(137,10,137).
The only difference in the three colors is the green level. The green level 10 is arbritarily selected to get the maximum contrast between the two stripes. We can't easily use a level less than 10 because the sRGB spec calculates low pixel values differently than the rest of the pixel value range. So all we want is a green level that, when combined with the 10 level, will appear to have the green level 137 used in the grey.

One may think to get a mean 137, we would need a combination 10 and 264. But as mentioned above, we can't apply math to pixel values, they need to be converted to luminance levels. So we plug 10 and 137 into the gamma-power function and get 0.0030 and 0.2502, respectively. The luminance we need to combine with 0.0030 to get a mean of 0.2502 is thus 0.4973. So we simply invert the gamma-power function and apply 0.4973 to get a pixel value of 187. So the color we need to put next to rgb(137,10,137) for the two together to appear as rgb(137,137,137) is rgb(137,187,137).

As long as your display is set to the sRGB spec, the two stripes together will appear as grey. But if your display gamma is too high, say 2.4, everything will look too dark, and the combination of green 10 and 187 will appear to be equivalent to green 140, not 137, giving the test stripe a green cast. Conversely, if your display gamma is to low, say 2.0, everything will look too light, and the combination of green 10 and 187 will appear to be equivalent to green 133, not 137 giving the test stripe a magenta cast. (Magenta is the opposite of Green in the RGB-CYM color wheel.)

This test image was fairly difficult to construct, but it's so easy to use! There are other color stripe gamma test images out there that use the same principle, but nearly all of them only test the gamma at one to three brightness levels. There's only one other test image that uses the entire range, but it doesn't model the sRGB spec, only the more simplistic power function. This is the only test image that tests the entire brightness range to match the sRGB spec. It is the only test with a scale so you can see exactly which pixel values are out of spec. It is possible for someone so inclined to build a custom lookup table that corrects for departures from spec along the brightness range that most graphics cards inevitably have. The only thing more accurate is the hardware sensors which attach to your screen and exactly calibrate your display. These cost a fair bit of money, this test image is free!

The Digital Dog image is Copyright © 2009 Andrew Rodney. My Gamma Test Image is Copyright © 2012 Glenn Messersmith. Creative Commons License It is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 3.0 Unported License.