Gamma revisited

[mikesan]

Not long ago I was having a beer with a fellow amateur photographer and he brought up the subject of "gamma" as it applies to digital imaging. He was stimulated (and puzzled) by a long contentious argument in an internet photo forum. Between us we don't command a wealth of information nor an in-depth understanding of the subject, beyond what one finds at numerous internet sites. What I mean to say is, I have read many of the most frequently cited articles, (among which the quality varies from clearly flawed, to murky, to comprehensible but incomplete), and there are still obvious gaps in my understanding.

Most recently I have read the tutorial published by CinC "Understanding Gamma Correction". It ranks as one of the best, but I still come away with the feeling that there are some very obvious unanswered questions, at least in my mind. A search of this site with the key words "gamma correction" shows that I am not alone. There I found many good questions and a few clear and concise answers. Still I remain unsatisfied.

I should state at the outset that I have a reasonably good background in math and science, but no particular knowledge of digital imaging or any of the associated diciplines. Given that limitation I set out to construct a tutorial (only for my own edification) based on a few accepted facts and some reasonable assumptions.

I would like to submit this explication to the readers of this forum for the purpose of obtaining critical feedback and especially corrections, when I have erred in my assumptions or conclusions. I hesitate to proceed with this project, first because I am very much a newcomer to this site, and also in that I am concerned many may find it not an acceptable use of the forum or a waste of your bandwidth; It may require a few lengthy posts. So I post this preamble to "test the water", and welcome your comments.

[dje]

I for one would be quite happy to see your tutorial. I have seen many instances of mis-information being repeated over and over on the web to the point where it seems it must be right. There is no harm in challenging such thoughts.

Dave

[Richard Lundberg]

Gamma as it usually used actually means gamma correction. Displays are commonly assumed to have a gamma of 2.2, and in fact LCDs are trimmed to that, but CRTS can be way off. Gamma refers to the exponent of the voltage(or current) to luminance transfer function of the display. This must be corrected to allow the eye's own transfer function to see images relatively naturally. As implemented, the curve is only accurate in the midrange for simplicity and to avoid problems at low luminance and high luminance. The low end and high end are linear. In effect, gamma correction is used to emphasize or de-emphasize the mid tones. "Gamma 1.8" generally is used when print is final output to compensate for the response of photo papers.

[Mark von Kanel]

HI MIke?

Firstly can you please enter your real name and location so we know a little more about our tutor?

No information given on here is a waste of bandwidth.... Some of it may be found to be incorrect initially but more often than not is corrected after lots of discussion and debate, So i look forward to your tutorial. Can i make one request from you and the other members contributing?

Please keep it all in plain English.....I know that jargon and technicle terms are going to be a nessasarry factor, but if you use one of these terms please take the time to explain what it means first! Already we have terms such as LCD, CRT, Gamma correction and a scale applied to said correction all of which haven't been adequately described and put into context (not picking on you Richard )

All to often these discussions get too technical all to quickly and the good information is hidden in paragraphs of "LINGO". For once lets see if we can discuss a technical subject and keep it understandable for ALL of the readers interested.....

Remember KISS Keep It Simple Stupid!!

Now im off to find out what Gamma correction is

[Simon Garrett]

I agree with you that many descriptions of gamma are inadequate, and some are downright wrong. I'd be interested to see yours. Any new description welcome, but be prepared for the experts here to tell you in no uncertain terms if they think you've got it wrong!

Most recently I have read the tutorial published by CinC "Understanding Gamma Correction". It ranks as one of the best, but I still come away with the feeling that there are some very obvious unanswered questions, at least in my mind.

The one on Cambridge in Colour is pretty good, I reckon (I've just read it). IMHO one point needs particular emphasis, and one point could be added:

• The point to emphasise: as the tutorial says, the overall system gamma (from camera to monitor output) must be 1. Yes, our eyes have non-linear response, but the same non-linear eyes see the original scene and the monitor, and the two won't look the same unless the overall gamma is 1. This point is often misunderstood and wrongly explained.
• The additional point: the tutorial talks of a gamma being an exponential function. Quite correct, but a gamma function is just one type of Tone Response Curve, and the TRC of images and devices is not always a gamma function. sRGB uses a complicated curve that approximates to a gamma of 2.2, but differs near black. Monitors and other physical devices often have response curves significantly different from gamma functions. The CiC article implies this talking of "Display gamma", but I think this point is worth adding (or making clear).

There are two main reasons for applying a gamma curve (or other TRC), as the CiC article says:

1. to provide more perceptually uniform coding levels, which means we can encode in a smaller number of bits (e.g. 8 bits). Not so important with 16-bit encoding.
2. to compensate for a non-linear TRC in a device such as a monitor (by applying a TRC opposite to that of the monitor)

The purpose is not to adapt the image to our non-linear perception - except indirectly in point (1) above.

In view of the comment in my first para, I shall now don a hard hat and dive for cover!

[mikesan]

HI MIke?

Firstly can you please enter your real name and location so we know a little more about our tutor?

Done.

I do not see myself as a tutor to anyone other than myself. However, having done some teaching in the past, I have learned that one way to be sure I understand a subject well is to attempt to teach it to someone else.

And thanks for your other suggestions. Simplicity is high on my priority list.

Since my "toe in the water" post did not elicit any negative responses, I will begin preparing the first episode. Stay tuned.

[ajohnw]

Personally I feel that a lot of the "information" on gamma is too scientific especially as it was introduced as a fudge factor. It better to understand what it is about.

The fudge is pretty simple really. Using the term EV's or stops loosely an 8 bit system might go 1,2,4,8,16,32,64,128, and then 255. The gradation steps get bigger as the levels get higher. In real light level terms visually they don't go like that at all. Gamma redistributes this so that effectively there are more stops visually but only after a fashion. In real terms it provides better grey gradation and also alters the rate of change at the dark end of things. The wiki offers a pretty good explanation including a scale which illustrates what is happening and an interesting test image - if clicked on. They also use the term subjective - good for them. At one time they did mention how many stops the range represents but that is probably debatable.

There are all sorts of odd comments about. Gamma reduces the number of stops available along with maths is my favourite. Simplifying how digital monitors do it is another. Using levels to adjust where mid tones lie adjusts gamma is another - in a sense that's correct but the gamma of the display isn't changing.

John
-

[mikesan]

A NOTICE: to those interested in my project.

It is apparent to me that anything more than a superficial treatment of this subject will unavoidably require some mathematical explanations. I shall try to keep the level no more complicated than high school algebra. For my own purposes I find that seeing a detailed solution of math problems is more informative than just stating a few equations. The use of an Excel spreadsheet is very useful for this purpose. So what I wish to do is attach one or more Excel spreadsheet files to future posts. Of course this will require that readers have access to Microsoft Excel on their computer. At this point I am assuming that the majority of those interested do have this software.

I solicit your reaction to this idea.

[dje]

A NOTICE: to those interested in my project.

It is apparent to me that anything more than a superficial treatment of this subject will unavoidably require some mathematical explanations. I shall try to keep the level no more complicated than high school algebra. For my own purposes I find that seeing a detailed solution of math problems is more informative than just stating a few equations. The use of an Excel spreadsheet is very useful for this purpose. So what I wish to do is attach one or more Excel spreadsheet files to future posts. Of course this will require that readers have access to Microsoft Excel on their computer. At this point I am assuming that the majority of those interested do have this software.

I solicit your reaction to this idea.

No problem here, I'm starting to get intrigued Just present it however you want Mike, you don't need to seek approval.

Dave

[mikesan]

No problem here, I'm starting to get intrigued Just present it however you want Mike, you don't need to seek approval.

Dave

Dave-
Thanks for your support. However it would do little good to attach Excel spreadsheet if a significant number of readers to not have the software to read them. I await further feedback.

[Stagecoach]

I look forward to seeing your article Mike and fire away with the Excel spreadsheet.

Grahame

[dje]

Dave-
Thanks for your support. However it would do little good to attach Excel spreadsheet if a significant number of readers to not have the software to read them. I await further feedback.

Mike I think most of the people who might be interested would have access to Excel but I could be wrong ! There will of course be plenty of people on this site who don't want to delve into this level of detail but they don't have to read it.

Dave

[Mark von Kanel]

No problem here, I'm starting to get intrigued Just present it however you want Mike, you don't need to seek approval.

Dave

+1 but be prepared for some very stupid questions!

I read the tutorial on heare about it and my eyes glazed over! i found myself asking why do i need to know this? all knowledge is useful but at the end of the day if i take a picture and process it on my calibrated screen and print it on my profiled printer and it looks good and i get the same results off any of my machines.... do i realy need to know???

Just being devils advocate

[Mark von Kanel]

Theres always open office Mike which is free and although im not certain i assume it will read excell documents. Post on Mike!

[Simon Garrett]

A NOTICE: to those interested in my project.

It is apparent to me that anything more than a superficial treatment of this subject will unavoidably require some mathematical explanations. I shall try to keep the level no more complicated than high school algebra. For my own purposes I find that seeing a detailed solution of math problems is more informative than just stating a few equations. The use of an Excel spreadsheet is very useful for this purpose. So what I wish to do is attach one or more Excel spreadsheet files to future posts. Of course this will require that readers have access to Microsoft Excel on their computer. At this point I am assuming that the majority of those interested do have this software.

I solicit your reaction to this idea.

Hmm... Make sure you don't hide the wood in the trees. The concepts of gamma really are not complicated and don't need much maths to explain, IMHO. The reason that many of the articles about gamma aren't very good is (IMHO) that the authors don't understand it properly, not because they haven't used enough maths to explain it.

I'd be really intrigued to know what will "unavoidably require some mathematical explanations" beyond those in the CiC article. I wonder if graphs and diagrams would explain it better than a spreadsheet? To me, a table of figures doesn't say much. It's when one can visualise the figures in some way that they take on meaning (as in the tone wedges in the CiC article).

If you are going to explain some of the maths of tone mapping, perhaps that should be after you've explained what I think are simple concepts in more simple terms.

However, I'm looking forward to your article.

Just my thoughts!

[Mark von Kanel]

Hmm... Make sure you don't hide the wood in the trees. The concepts of gamma really are not complicated and don't need much maths to explain, IMHO. The reason that many of the articles about gamma aren't very good is (IMHO) that the authors don't understand it properly, not because they haven't used enough maths to explain it.

I'd be really intrigued to know what will "unavoidably require some mathematical explanations" beyond those in the CiC article. I wonder if graphs and diagrams would explain it better than a spreadsheet? To me, a table of figures doesn't say much. It's when one can visualise the figures in some way that they take on meaning (as in the tone wedges in the CiC article).

If you are going to explain some of the maths of tone mapping, perhaps that should be after you've explained what I think are simple concepts in more simple terms.

However, I'm looking forward to your article.

Just my thoughts!

And im looking forward to your input on the article!!! just for us KISS fans

[MrB]

MS Excel Viewer:

http://www.microsoft.com/en-us/downl...ils.aspx?id=10

Cheers.
Philip

[ajohnw]

Theres always open office Mike which is free and although im not certain i assume it will read excell documents. Post on Mike!

The right version of open office will. Not the one on the Apache site.

John
-

[TonyW]

I just had a look at the CiC tutorial and did not find it particularly clear. I would be glad to see an alternative description and am happy with Excel.

[ajohnw]

I just had a look at the CiC tutorial and did not find it particularly clear. I would be glad to see an alternative description and am happy with Excel.

I suspect the confusion is due to mention of gamma as a number rather than what it is. It's a transfer function. It takes some input and changes it to some other value but with the characteristic on output devices that it leave pure black 0,0,0 and pure white 255,255,255 alone and only alters values in between. A gamma of 1 is no change. A gamma of 2.2 corresponds to a power law of 1/2.2 or 0.4545.

There is a tendency to think that all of this goes back to CRT technology as they have a natural gamma but that isn't the case. The reason for it's use is to make linear systems "see or show" in a similar fashion to the human eye that has a logarthmic response to light levels, another power law. The wiki gives a good run down but probably introduces too many complications due to gamma compression, using gamma correction to change an image that has gamma into a linear one that hasn't and then applying gamma again when it's actually seen and etc. Some of this sort of thing is done to give a more pleasing result to viewers.

http://en.wikipedia.org/wiki/Gamma_correction

In terms of sRGB it's probably better to look at this one

http://en.wikipedia.org/wiki/SRGB

But as usual there are mistakes - CRT monitors generally needed trimming to have a gamma of 2.2. Some had a knob to adjust it and I'd guess all decent ones had internal circuitry to trim it to the desired value. Just like CRT TV's. On the other hand they do show the transfer function that sRGB uses. The curve will always have the same shape on other colour spaces that use a gamma of 2.2. The whole thing is something HP came up with when Microsoft screamed help but in real terms it's been about for a long time for the same reasons.

Really for many people on here it's all no different to using curves in a PP package. A straight line leaves it alone, gamma is 1. Change it and the gamma of the image is being changed but personally I feel that using the word gamma here is probably adding to confusion even though it's exactly the same basic process - contrast is being altered. In the case of gamma to suit the human eye and in the other case to get a nice picture.

John
-