Making fine prints in your digital darkroom
Background to monitor calibration and gamma
by Norman Koren

Site map/guide to tutorials
Contact | News

Making fine prints in your digital darkroom
Understanding image sharpness and MTF
Image galleries / How to purchase prints
Photographic technique
Image editing with Picture Window Pro

A simplified zone system
Digital vs. film
updated Feb. 15, 2005
View image galleries

Google

Search WWW Search www.normankoren.com

 

Table of contents

for the Making Fine
Prints series

Getting started | Light & color | Scanners
Digital cameras | Printers | Papers and inks
Monitor calibration and gamma
Background to monitor calibration and gamma
Why a new gamma chart?
Three level chart | Three color chart
Luminance steps | Timo Autiokari
Printer calibration | Scanning | Basic image editing
Black & White | Matting and framing
Tonal quality and dynamic range in digital cameras

Color Management: Introduction | Implementation
Profiles with MonacoEZcolor | Evaluating profiles




for Image editing with
Picture Window Pro
Introduction | Making masks
Contrast masking
Tinting and hand coloring B&W images
Example: Sunset, Providence, Rhode Island

In this page we cover some background (reference) areas related to gamma. We discuss why a new chart was needed (many old ones had an error), then we present two additional charts: one for three luminance levels and one for three colors (R, G, B). We illustrate what luminance steps look like for different gamma settings. We moved the Praxisoft WiziWYG instructions to this page: since I put up QuickGamma, WiziWYG is no longer my choise for monitor calibration. Finally I discuss pages by Timo Autiokari, which contain some dubious advice on gamma.

Why a new gamma chart?

Most gamma charts consist of an arrangement of black and white patterns and solid tones. You estimate gamma by viewing the chart from sufficient distance (typically one to two meters from the monitor) so you don't see the pattern details, then either locating the position where the average luminance of the pattern matches the solid tone or operating a slider to obtain a match. Many of these charts use checkerboard patterns, typified by the pattern on the left: a portion of the gamma chart supplied with Epson scanners, enlarged 3x to make the pattern plainly visible. This type of checkerboard pattern gives an erroneous estimate of gamma! I owe a debt of gratitude to Pete Andrews for pointing this out.

The problem is the risetime of the video card and CRT monitor. In most cases the brightness cannot change from pure black to pure white in the short distance of one pixel (or even two or three). Since CRTs are scanned horizontally, abrupt vertical features are affected.

The situation is illustrated in the box on the right, which consists of four quadrants, each of which contains equal amounts of Black and White. 1: vertical alternating B and W. 2: horizontal alternating B and W. 3: Vertical B and W alternating every second line (same as 1 magnified 2x). 4: Horizontal B and W alternating every second line (same as 2 magnified 2x).

If risetime were not a problem, all four quadrants would appear equally bright when viewed from a distance. But on most CRT monitors, quadrant 1 (alternating vertical lines) appears quite different from 2 (alternating horizontal lines)-- usually darker. Quadrants 2 and 4, both containing horizontal lines, should be similar. Conclusion: patterns with vertical features are sensitive to risetime and cannot be used reliably to estimate gamma. Liquid crystal displays (LCDs) don't have this problem, but most have another, worse, problem-- sensitivity to viewing angle. For more details see Pete Andrews' Monitor calibration and Gamma assessment page.

I was dismayed to discover that my monitor's gamma, which measured around 2.1 with the old chart, was closer to 2.6. I corrected it using my video monitor's software (illustrated above ). I'll be recalibrating my printer and editing many images in November. Because of the gamma error I set Contrast to +12, well above the default setting (0). I'll undoubtedly reduce it. Annoying and a little embarassing, but the end result will be improved print quality.

Three level and three color charts

Three level gamma estimation chart

The three level chart on the right enables you to estimate display gamma for three normalized luminances: 0.25 (left), 0.5 (middle) and 0.667 (2/3) (right-- 0.75 was too light to be legible). Gamma is estimated for each zone by locating the position where the average intensity across the zone is constant. The corresponding gamma is shown on the right. The section for luminance = 2/3, on the right, is difficult to read.

I use this chart mainly as a diagnostic tool-- to see if gamma is consistant at different tonal levels. If your monitor is working well, gamma should vary by no more than  ±0.1 between the zones. For normal gamma estimation I use the combined gamma/black level chart, on the previous page. The solid areas are calculated from the equation,

pixel level = 255*luminance(1/gamma).
Three color gamma chart

A gamma chart with separate R, G and B patterns is shown on the right. It could be useful if you need to adjust gamma separately for each color. The blue pattern is dark-- difficult to read. I don't find this chart very useful for diagnostics; if gamma is different for different colors, you'll see color variations in the gray chart. But it could be useful in adjusting gamma in special cases, like my ATI Radeon video card, where the software doesn't allow the gamma adjustment to be coupled for the three colors.

Steps in pixel level and luminance

The following table displays 21 luminance level steps, from minimum to maximum, in two different ways. In the top row the pixel levels are evenly spaced. In the bottom row the luminance is evenly spaced (linear stepping) if monitor is calibrated for gamma = 2.2, The upper number in each box is the normalized pixel value (pixel level/255). The lower level is the normalized luminance for a monitor with gamma = 2.2. There are approximately 12.75 luminance steps per box (255 total). [These tables will not display correctly if you are using Netscape and the Always use my colors, overriding document box is checked (under Edit, Preferences, Appearance, Colors). You must uncheck it.]
 
21 even steps in pixel level (0-255; 0-1 in steps of 0.05, normalized)
Normalized pixel level on top; screen brightness for gamma = 2.2 below.
0
.05
.001
.10
.006
.15
.015
.20
.029
.25
.047
.30
.071
.35
.099
. 40
.133
.45
.173
.50
.218

MID

.55
.268
.60
.325
 .65
.388
.70
.456
.75
.531
.80
.612
.85
.699
.90
.793
.95
.893
1
1
00.
0
.256
.05
.351
.10
.422
.15
.481
.20
.533
.25
.579
.30
.621
.35
.659
.40
.696
.45
MID

.730
.50

.762
.55
.793
.60
.822
.65
.850
.70
.877
.75
.904
.80
.929
.85
.953
.90
.977
.95
00.
1
1
21 even steps in screen luminance for gamma = 2.2 ( 0-1 in steps of 0.05, normalized)
Normalized pixel level on top; screen brightness for gamma = 2.2 below.


Note the large relative luminance steps at low levels in the lower row, particularly between the first and second box. This is what you would get for evenly stepped pixel levels with monitor gamma = 1, resulting in banding in dark areas (below). In the top row, the relative luminance steps at low levels are relatively small; the second box should be perceptibly lighter than the first, but still very dark. On the whole, gamma = 2.2 is more perceptually uniform than gamma = 1. Gamma = 1.8 (the value used in Macintosh) has near optimal perceptual uniformity; it has a slight edge over gamma = 2.2.

The middle (11th) box in the upper row (pixel level 127) has a normalized luminance of 0.218. In photography, the 18% gray card is considered "middle gray." Reflective exposure meters are designed to give correct readings when the gray card is used; it is regarded as being close to the average density of the "average scene." The flip side of the gray card is 90% reflectance white. The normalized reflectance for middle gray is therefore 18/90 = 0.2: very close to the 0.218 luminance of the middle box-- pixel level 127-- for gamma = 2.2.

The following table zooms in on the above table to display the worst case banding. The top row, for highlights, has evenly spaced pixels from 236 to 255 in steps of 1. The bottom row, for shadows, has evenly spaced luminance steps when viewed with gamma = 2.2, the normal setting for Windows and the Internet. It illustrates the banding that takes place with gamma = 1.
 

Worst case banding examples
Light areas: 20 even steps in pixel level (236-255 in steps of 1)
Pixel level on top; screen brightness for gamma = 2.2 below.
255
1
W
236
.843
237
.851
238
.859
239
.867
240
.875
241
.883
242
.891
.243
.899
244
.908
245
.916
246
.924
247
.932
 248
.941
249
.949
250
.957
251
.966
252
.974
253
.983
254
.991
255
1
W0
 
Blk
0
21
.004
28
.008
34
.012
39
.016
43
.020
46
.024
50
.027
53
.031
56
.035
59
.039
61
.043
64
.047
66
.051
68
.055
70
.059
72
.063
74
.067
76
.071
78
.075
Blk
0
0
Dark areas: 20 even steps in screen luminance (in normalized increments of 1/255),
illustrating integral pixel steps (0,1,2,3,...) for gamma = 1 (when displayed at gamma = 2.2)
Pixel level for gamma = 2.2 on top; normalized screen brightness (n/255) below.


I find the highlight banding to be barely preceptible, if at all. But the shadow banding-- what you would get with gamma = 1-- is obvious and unacceptable. There would be no visible shadow banding with gamma = 1.8 or 2.2.

.
 Timo Autiokari and AIM (Accurate Image Manipulation)
Timo Autiokari's Accurate Image Manipulation for Desktop Publishing website is beautifully designed, rich in content, and seems well organized at first glance. Timo thinks for himself; he doesn't follow the herd.

But his page contains significant misinformation. He states that image editing introduces serious errors unless gamma is set to 1. And indeed, problems can arise in operations that involve strongly contrasting adjacent pixels.

Blurring, which involves averaging neighboring pixels, is a good example. If you blur two adjacent pixels with values 1 and 255, you get pixel values around 128, which appear darker than the average of pixels 1 and 255 if gamma is larger than 1. (This is the basis of the gamma chart, above.) This effect causes visible errors with extreme sharpening and heavy blurring (visible in Timo's examples), and rotation (see Helmut Dersch's page on interpolation; alternate site), but there are no problems with operations on individual pixels such as color and tonal adjustments. And none with reasonable (not extreme) sharpening-- only with severe oversharpening with exaggerated edge contrast.

For the most part, Timo is wrong about gamma. The primary reason for rejecting his assertion is the banding in shadow areas with gamma =1, illustrated above. This banding would be visible on prints. You are best off sticking with your system's standard gamma (2.2 for Windows, 1.8 for Macintosh) for image editing.

Timo doesn't comprehend the simplicity of the gamma relationship, luminance = (pixel level/255)gamma. In the Introduction to System Calibration, he states,

"Remembering that all the hues on the screen are additive mixtures of the red, green and blue primary colors it is easily understood that gamma produces hue-shift also when these tri-color codes are edited."
Absolute nonsense! Though his statement would be correct if RGB data were uncorrected for gamma. In practice, the same gamma is assumed for each individual color in an RGB file; there is no hue shift when it is printed or displayed. In editing operations that affect color and brightness, RGB images are temporarily converted into a format such as HSV or HSL where hue, saturation and lightness (or brightness) are decoupled. Again, there is no hue shift, though these operations are performed with best accuracy in 48-bit color.

Timo assumes that 2.5 is the standard gamma for PC's. Wrong! The de facto standard is 2.2. But Timo insists on using 2.5 for examples throughout his page. 2.5 is the average gamma of an uncorrected monitor, but I've seen values between 2.2 and 2.8. That's why gamma adjustment is important.

The more I grapple with Timo's "logic," the more frustrated I get. His examples look impressive from a distance, but they are obtuse and often miss the point entirely. Charles Poynton has responded to Timo's savage, emotional attack (which I find to be so mathematically muddled as to be unreadable) with a cautionary tale. Poynton knows his stuff. As much as I like mavericks, I must side with the establishment on this issue. See also, Timo's technical argument | Timo and linear coding| Linear and nonlinear coding | Gamma FAQ.

Back to Monitor calibration and gamma
.


Images and text copyright (C) 2000-2013 by Norman Koren. Norman Koren lives in Boulder, Colorado, where he worked in developing magnetic recording technology for high capacity data storage systems until 2001. Since 2003 most of his time has been devoted to the development of Imatest. He has been involved with photography since 1964.