Measuring monitor contrast ratios with a camera

It is nearly impossible to accurately estimate the contrast ratio of a monitor by eye, and the manufacturers are often inflating the numbers to unrealistic values. However, with a digital camera and this web page, it is quite easy to check it yourself. It is most accurate with a camera that gives you some manual control over exposure time and the like, but it will work even with a point-and-shoot camera.

The idea is simple: take a picture of a white area and of a black area on the screen. The brightness of the picture, combined with camera settings such as ISO, F-number, and shutter time tells us how much light comes off the screen in either case. For the most accurate results, all camera settings should be identical, except for the shutter time, which is chosen such that both pictures end up as a middle grey.

Measuring the contrast ratio with a camera. Note how the camera is held close to the screen.

These are the minimum settings:

• White balance: set to "cloudy" or "daylight". (Do not use the default "auto" setting).
• Flash: OFF.
• Zoom: use the wide-angle setting, that is: fully zoomed out. The analysis software on the server will crop the image to the equivalent of viewing the screen from a distance of 1.5 times the screen diagonal. (If you don't want to use the image upload option, then zoom to about 2×)

Cameras vary in how easy it is to adjust the exposure. A good starting point seems to be ISO 400, F 2.8, and shutter times of 1 and 1/400 second. (For laptop screens, use 1/4 and 1/400 second). Always keep the same ISO and F number for the two images, and vary the shutter times if necessary to get two images with approximately the same mid-grey level (2.5 stops below saturation). Also try to keep the shutter times between 2 and 1/500 second. Set the camera to a low-resolution mode; it will reduce image noise. Also don't try to focus on the screen, because we are not interested in seeing individual pixels.

If the above is cryptic, you can try the following procedure:

• Camera mode: set to aperture priority ("Av"), or manual ("M") if that is not available.
• ISO value: set to 400.
• Exposure value adjustment (EV): start at 0.
• Aperture value (F-number): set to the lowest possible value, e.g. 2.8.
• Image resolution: low, for example 640x480.

Then proceed with the test as follows.

1. Make the environment as dark as possible.
2. Set the shutter time to about 1 s (or 1/4 s for most laptop screens). Put the camera very close to the black area below, at a straight angle, and take a picture of the black area. Don't worry about warnings about camera shake and autofocus. (If the camera has manual focus control, set it to infinity, ∞) If the image is very dark, increase the shutter time (or change the EV adjustment) and take a new picture to get a mid-grey image.
3. If the black level of your monitor is worse close to the edges ("backlight bleed"), then you may repeat the procedure for the borders of the screen. Make sure that you don't photograph the light from the menu or task bar and such.
4. Decrease the shutter time to about 1/400 s. Make a picture of the white area. Change the shutter time or EV adjustment if the image is (almost) saturated white or very dark rather than mid-grey.
5. Upload two representative JPG files in the form below. The result page will warn you if you need to change your camera settings. Optionally, you can add a third JPG file taken at the edge of the black screen.

Alternative method

If you don't use the upload form, you can proceed as follows.

1. Review the pictures. If either of the two is vastly underexposed or overexposed, try changing the exposure time or EV adjustment. If that doesn't work, you can change the aperture F number and the ISO number, but these two should be identical for the black and white photographs. That means: don't shoot the white area at ISO 100 and the black area at ISO 800. Also, avoid extremely short shutter times (1/1000 or less).
2. Look at the pictures and if your camera has one, the histogram. For both the dark and the light pictures, locate the peak of the histogram. In the example image, the peak is at 2.5 stops. If the peak (measured at the base) is wider than about 1 stop, I would suggest to read the value 0.5 stops from the right edge. Alternatively, enter the RGB values that are in the resulting image file (see below for hints).
3. Enter the numbers in the form below. (This uses javascript. It should work with Firefox 1.5/2.0, Opera 9, and even Internet Explorer 6.)

Quantity	Black image	White image	Explanation
Exposure time			Best between 0.5 and 1/500 s.
F-number			Preferably equal
ISO value			Preferably equal
Histogram peak RGB values			Need 0.1 or better accuracy. RGB values in the range 1–255. sRGB   Adobe RGB   (see below)
Luminance	??	??	Read the explanation below!
Your contrast ratio
Contrast ratio	??

Background

Using RGB values

Picking RGB values in GIMP.

Picking RGB values in MS Paint.

If your camera does not have a histogram, or a histogram without a scale, and you don't want to upload your photos for some reason, you can open the JPG image file in an image editing program such as the GIMP or MS Paint. Of course, there are plenty of other programs, too, but these are easy to get. This method is more accurate than looking at the histogram, but also more time-consuming. The RGB values are converted into luminances (relative to image saturation) according to the sRGB standard [Wikipedia] or Adobe RGB standard [Wikipedia]. Most consumer-grade cameras use sRGB.

Especially with the black image, there might be a considerable difference between the center and the edges. In that case take the RGB values at the lighter parts of the image.

a. GIMP procedure

GIMP is available for all common platforms (Linux, Mac, Windows) and is free. [download]. Select the color picker tool (1; it looks like an eye dropper). Set the "sample average radius" to 50 or so and click at a representative part of the image (2). Depending on the GIMP version, you will get a popup with the RGB values (3), or you can click on the currently selected color (4) to get a window with the RGB values (5).

b. MS Paint procedure

MS Paint is included in Windows (it's under the Accessories menu). Select the color picker tool (1; it looks like an eye dropper). Click at a representative part of the image. Then go to Colors (2) → Edit Colors → Define Custom Colors (3) and read the RGB values (4). Repeat this a few times and calculate the average RGB values.

Contrast ratio accuracy

I guess that it's fairly evident how this test works. The result should be fairly reliable if your camera can show a histogram. There are a number of error sources:

• Inaccuracy of reading the histogram peak position (±0.1 stops means ±7% in the contrast ratio).
• 
  On some screens, the black level has a strong angle dependence, such as in the image on the right. The center of the photo, corresponding to viewing the screen straigth-on, is fairly dark (3.8 stops, or 0.23 cd/m2), which results in a 437:1 contrast ratio. However, the edges, corresponding to a normal working distance from the screen, are much brighter (1.5 stops, 1.16 cd/m2), which results in a 89:1 contrast ratio. I would say that the latter (worse) number is more realistic, since you usually don't watch the screen exactly straight on.
• Inaccuracy of the F-numbers if they were different for the black and white images. The camera may tell you that it is 5.0, but it could well be 4.8 or 5.2. That's why it's better to use the same camera settings for both images, such that this error is canceled.
• Inaccuracy of the ISO numbers. The camera sensitivity is often not proportional to the ISO setting.
• Inaccuracy of the shutter time. Especially a problem for very short shutter times, e.g. 1/1000 seconds or shorter.
• Sensor noise. At high ISO speeds, cameras typically produce noisy pictures, which blurs the histogram, or requires averaging over many pixels when using the RGB method. Setting the camera resolution to e.g. 640x480 lets the camera do the averaging for you (on a 4 megapixel camera that means averaging over 13 pixels). The camera averaging acts on the raw sensor data, which is more accurate than when you do it on the already-processed JPEG data.
• Focusing. It's best to keep the camera completely out of focus, such the camera doesn't see the individual pixels on your monitor.
• The backlight of some LCD monitor flickers at relatively low rates (e.g. 300 Hz), which may affect the results for short shutter times. That's why you should take the average of several readings in such cases.

Luminance accuracy

The form also shows the absolute luminance as calculated from the data entered. All numbers should be known and not guessed or this number will not be meaningful. And even then, the real ISO numbers of cameras may vary considerably from one camera model to the next, so don't take this number as an absolute truth.

The calculation assumes that the sensor or RGB image saturates at a luminance

where f# is the aperture number, and t is the exposure time. This equation uses the "standard output sensitivity (SOS)" definition of the digital ISO value. [see Wikipedia]. From some experiments with a calibrated luminance meter and various cameras, it appears that the actual sensitivies can differ considerably from this equation, and that it is not always consistent on the same camera at different "ISO" settings. Here are the numbers for a couple of cameras:

Contact

---

• Home
• LCD test
• LaTeX
• Linux
• Web
• Misc

---

[Email]