How far to the right are you?

[chauncey]

Attention to minutiae, both in theory and in practice, is what distinguishes the good photographer from the point-and-shoot guy.

I might submit that the definition of a "good photographer" would be akin to defining "what is Art"...
defining either would be a task bordering on the absurd.

The ability to occasionally cough out a pretty picture, however it is achieved, would rank higher on my
priority list than would that of being a good photographer, whatever that means.

[Manfred M]

And that's where the last part of this thread was about. There is no reason why that should be. On contrary, blinking highlights should be and are less in JPG. Unless, and that's not mentioned jet, it's the thumbnail size that's making the difference. Reducing the size in pixels might have some influence.

George

I think this is going to be rather automatic due to the compression algorithms associated with (a) reducing the base 12-bit or 14-bit RAW data down to 8-bit used in jpegs and (b) will be further altered by the lossy compression process.

We can only speculate as to how the compression algorithms were implemented as we do not have access to the source code in either our camera firmware or editing software. I quite agree though, depending what file (thumbnails or otherwise) are use to calculate blown out highlights are going to drive the results; accuracy is always a function of data quality; garbage in = garbage out.

That being said, even when I get areas indicating highlights are blown out when running a RAW file through ACR, I can usually recover detail by moving the white and / or highlight sliders to the left. That tells me that there is highlight detail in the RAW data, but compressing the data so that it can be displayed on the computer screen is likely causing the highlight warnings.

[ajohnw]

I might submit that the definition of a "good photographer" would be akin to defining "what is Art"...
defining either would be a task bordering on the absurd.

The ability to occasionally cough out a pretty picture, however it is achieved, would rank higher on my
priority list than would that of being a good photographer, whatever that means.

Some of this thread has made me laugh mainly due to shooting EM's. On the latest one I sometimes find it's better to shoot to the left. Not extremely but deliberately under exposing as they don't according to what I have read but not checked leave very much extra in raw. circa 1/2 stop I believe. If the dynamic range there is suitable why not. Noise relates to that as well. I also have a fair idea when I can use 25600 ISO. It's easy to shoot without highlight clipping in these cameras. I'm not saying look what I have only that it's different. I retain the comments I have made in respect to other cameras for if I ever use my Nikon or Canon's again. Basically they are all different even model to model and it's best to take some time to learn what they do in practice. I probably will never fully learn how to exploit the Nikon and if that way inclined might even go back to Canon as I have a fair idea what they do.

Your right few if anybody always produce great shots every time who ever they are.

John
-

[george013]

That being said, even when I get areas indicating highlights are blown out when running a RAW file through ACR, I can usually recover detail by moving the white and / or highlight sliders to the left. That tells me that there is highlight detail in the RAW data, but compressing the data so that it can be displayed on the computer screen is likely causing the highlight warnings.

Personaly I think a kind of posterization. The highest pixelvalue in 8-bits is 255, just 1 figure. That same pixel can have in a 12-bit division 16 values numbered 4080 til 4096. Viewing the image doesn't show a difference, it's all 8-bit. But editing gives a difference. That's the way I see it.
To me editing both a RAW-file as a JPG-file happens on a rasterimage with the only difference the bitdepth. The RAW-data is only used to create a "master" rasterfile.
George

[DanK]

Regretfully, the reality behind those "minutia" distorts serious discussion, as people use terminology sloppily, talking about over-exposure and under-exposure and correct exposure, without thinking twice. When words are used in the wrong way, when one participator in a discussion uses words that are ill defined, sloppily, it is very difficult to partake. Therefore, it is not a minor thing, that there is a software control called exposure, as there is no software that can ever control exposure, except the software used in direct connection to the camera before exposure, be it firmware, tethered software or untethered.

I'm going to try to make peace here.

To put my cards on the table: I love technical details. I teach technical details. I certainly recognize that what seem like small and arcane details can matter.

But matter for what purpose?

My starting point, when I teach technical material, is to ask: what do these particular people need to know, to answer their pressing questions? So, go back to the OP. The fact is that many of the details in this thread are not needed for answering the OP's question.

So I see this thread as including two very different discussions. One is far more technical than the other. They are addressing different questions, which are relevant to different people. I won't disparage either part of the thread.

[ajohnw]

Personaly I think a kind of posterization. The highest pixelvalue in 8-bits is 255, just 1 figure. That same pixel can have in a 12-bit division 16 values numbered 4080 til 4096. Viewing the image doesn't show a difference, it's all 8-bit. But editing gives a difference. That's the way I see it.
To me editing both a RAW-file as a JPG-file happens on a rasterimage with the only difference the bitdepth. The RAW-data is only used to create a "master" rasterfile.
George

That's fine George but bear in mind that gamma remaps the number of bits to each stop. The same sort of thing is happening when a raw file is loaded*. The hidden bits are really being used as fractions of the display bits during PP so can be manipulated into the display. Tempted to add a comment about editing in 8 bit, I wont other than an aspect that has already been mentioned - a 1 bit change at that end often can't be seen visually.

Jpg compression is also pretty complex and can even be lossless. In practice 95% quality levels are rather near to that. 65% is a number sometimes used for web shots to keep bandwidth down while still producing decent images. Suggestions of 80% 85% aren't unusual in PP packages and these can still be reworked.

*If you want to see what happens when this isn't done it's an option in a application called dcraw. It might be in ImageJ as well but I don't think so.

John
-

[xpatUSA]

I would say that it does matter. As the erroneous designation of a software control is named to something no software can control, is not a minor thing, not at all minutia, because it clouds any serious discussion about the very subject, that regards exposure. Therefore, it is not minutia. It is an erroneous name that throws a smoke curtain over serious discussion.

Interestingly, during installation, Sigma Photo Pro loads an editable text file with all the button and slider labels in it. For example the 'save as' dialog has a cancel button in it. However the stupid button text says "Canceling" although it has not been pushed. Irritating, mis-leading and . . . . fixable!

So I could also go to their "Exposure" slider label and edit it to whatever I like, maybe "linear luminance transfer gain" (just kidding).

[george013]

That's fine George but bear in mind that gamma remaps the number of bits to each stop. The same sort of thing is happening when a raw file is loaded*. The hidden bits are really being used as fractions of the display bits during PP so can be manipulated into the display. Tempted to add a comment about editing in 8 bit, I wont other than an aspect that has already been mentioned - a 1 bit change at that end often can't be seen visually.

John, there are no hidden bits. The RAW-converter works with a 12-bits rasterfile in memory. To me this is the only time that the sensorinfo is used: creating the rasterfile with RGB pixel structure, using whitebalance and gamma.
Your screen is 8-bit. So corresponding pixels in the 12 bits range have to be rounded to 8 bits. Just for the screen. Editing happens with 12 bits. And that's where the so called extra information of a RAW-file comes from.
And also keep in mind that when somebody says that in a RAW-file is more info, that he compares that with the JPG, a 8-bits compressed file.

George

[xpatUSA]

Attention to minutiae, both in theory and in practice, is what distinguishes the good photographer from the point-and-shoot guy.

I might submit that the definition of a "good photographer" would be akin to defining "what is Art"...
defining either would be a task bordering on the absurd.

The ability to occasionally cough out a pretty picture, however it is achieved, would rank higher on my
priority list than would that of being a good photographer, whatever that means.

Whilst the need to defend your previous rant is understandable, I'm not certain that jumping my overly-simplistic sentence helps you very much in that regard. But, so as not to start a war, I will most obligingly re-phrase it for you:

Attention to minutiae, both in theory and in practice, is one means of improving one's photographic performance. Ignoring it and calling it "nonsense" is not.

[Stagecoach]

Attention to minutiae, both in theory and in practice, is what distinguishes the good photographer from the point-and-shoot guy.

One of the problems many people encounter when 'learning' something is what to take notice of, what's applicable to them and the value to them of what can be gained from new knowledge.

One of the obstacles always encountered is the credibility of the source and when I read comments like your's above Ted I will wonder if the 'technical' information offered is as 'loose' as the term "good photographer" in that sentence.

[ajohnw]

John, there are no hidden bits. The RAW-converter works with a 12-bits rasterfile in memory. To me this is the only time that the sensorinfo is used: creating the rasterfile with RGB pixel structure, using whitebalance and gamma.
Your screen is 8-bit. So corresponding pixels in the 12 bits range have to be rounded to 8 bits. Just for the screen. Editing happens with 12 bits. And that's where the so called extra information of a RAW-file comes from.
And also keep in mind that when somebody says that in a RAW-file is more info, that he compares that with the JPG, a 8-bits compressed file.

George

The fractional bits are hidden as far as the display is concerned George.

John
-

[george013]

The fractional bits are hidden as far as the display is concerned George.

John
-

Could you explain that?
George

[ajohnw]

Could you explain that?
George

Only as I already have done and only simply neglecting certain complications as far as real photographs are concerned but is possible.

Say 10 bit precision is mapped into 8 bit precision for display purposes. 0ne has a count of 1024 and the other 256, 4 times less so as far as the display is concerned each of the 10bits is a 1/4 bit that is visually hidden and can be made visible via PP, usually at the loss of some others of course.

So in principle various tonal ranges can be altered to fine limits. PP for instance might move something by up to any old number say 3 1/2 bits in 8 bit precision terms and 4 times that amount as far a the higher precision data that is there is concerned.

This might sound crazy but when mapping tones what is actually happening is that the rate of change/level is being modified - contrast and or brightness depending on what is done. A typical example where these fractional bits MAY be needed is when dark areas are brightened probably along with a contrast change to make them look as they should in the final image. That may mean that the 1/4 bits get stretched to the same levels as the 8bit bits or even further. This end of things is one of the main reasons for the use of 14bit D/A's as there are way too many bits at the highlight end but mapping and gamma also do something about that aspect. All down to the devices being linear and vision far from it.

John
-

[xpatUSA]

One of the problems many people encounter when 'learning' something is what to take notice of, what's applicable to them and the value to them of what can be gained from new knowledge.

One of the obstacles always encountered is the credibility of the source and when I read comments like your's above Ted I will wonder if the 'technical' information offered is as 'loose' as the term "good photographer" in that sentence.

Looks like we were typing at the same time, Grahame. I reworded the offending sentence in post #69.

[george013]

Say 10 bit precision is mapped into 8 bit precision for display purposes. 0ne has a count of 1024 and the other 256, 4 times less so as far as the display is concerned each of the 10bits is a 1/4 bit that is visually hidden and can be made visible via PP, usually at the loss of some others of course.

-

I don't understand anything of this.

George

[ajohnw]

I don't understand anything of this.

George

Sorry George can't help any further.

John
-

Well I suppose I can after a fashion. Mapping in this context means taking some value and translating it some fashion into another. In this area it''s done with a tone curve - a sort of graph with what goes in on one axis and what comes out on the other - 10bit preciion 0-1023 on one axis read off on the other axis as values between 0 and 255. PP steps can alter the shape of the curve.

John
-

[george013]

Sorry George can't help any further.

John
-

Well I suppose I can after a fashion. Mapping in this context means taking some value and translating it some fashion into another. In this area it''s done with a tone curve - a sort of graph with what goes in on one axis and what comes out on the other - 10bit preciion 0-1023 on one axis read off on the other axis as values between 0 and 255. PP steps can alter the shape of the curve.

John
-

John,
You don't read or understand my posts, and I hardly can read and understand your posts.

Once more the basics as I see.

The RAW-converter creates a rasterfile of 10,12 14 or whatever bit deep, depending on camerasetting.
Your editing is done on that let's say 12-bits rasterfile.
Your viewing is done on a 8-bit screen.
More info in a RAW-file is based on a comparison between a JPG and a RAW image. The difference is the bit-depth and the used JPG-compression. A JPG is 8-bit.
A JPG-compression should theoratical result in less highlight-clippings. I will do some experiments on that furher.
Different RAW-converters use different settings and thus giving different results.
The caluclation from x bit to y bit is (act.val.x/max.val.x)*max.val.y. Just staight on. You may choice what to do with the decimals, rounding or truncating.

George

[xpatUSA]

John,
You don't read or understand my posts, and I hardly can read and understand your posts.

Once more the basics as I see.

Let's see if we can discuss your basics one by one without fighting?

The RAW-converter creates a raster file of 10,12 14 or whatever bit depth, depending on camera setting.

Your editing is done on that let's say a 12-bit [per channel] raster file.

Most editors convert the raw file from the camera to a 16 bit per channel (bpc) working file, not 10,12,or 14. Adobe is known to use signed 16 bpc leaving 15 significant bits for editing.

Your viewing is done on a 8-bit screen.

Agreed, just for the purposes of this discussion.

More info in a RAW-file is based on a comparison between a JPG and a RAW image. The difference is the bit-depth and the used JPG-compression. A JPG is 8-bit. A JPG compression should theoretical result in less highlight-clippings. I will do some experiments on that further.

Introduction of JPEG compression brings an unnecessary complication into the discussion because there are many, many different types and levels of JPEG compression. As to the effect of JPEG compression on highlight clipping please publish the results of your experiments when completed, I would be interested.

Different RAW-converters use different settings and thus giving different results.

Yes

The calculation from x bit to y bit is (act.val.x/max.val.x)*max.val.y. Just straight on. You may choice what to do with the decimals, rounding or truncating.

By saying "the decimals", you are implying a floating-point calculation which is not used by all converters/editors. Having said that, your statement is about the same as John's graphic example of a transfer characteristic from one bit depth (X-axis) to another bit depth (Y-axis) but he stated further that the transfer characteristic is not necessarily linear. One example of this is Nikon's NEF compression where (for Nikon D50) the 12-bit camera data (4096 levels) is compressed to a non-linear (gamma-corrected) 683 levels.

Hope this is understandable . . .

[george013]

Ted,
With all respect, again you don't see the issue.
It's about the statement that there seem to be more info in a RAW-file than in a JPG and in-camera blinking happens earlier in a JPG then in a RAW.
If a RAW-converter works standard with a 16-bit workfile doesn't add anything to that. The camera gives a file with a bit-depth of 10,12,14 or whatever your camera is set on. And a JPG is 8-dit, without decimals.

For the others if they are interested, I did several comparisons in as well Capture as Rawtherapee. Choice a RAW-picture with blinkies light and dark, load it in the converter and write it back to file as a JPG. Then open it again and compare the two files and there blinkies. You have the same file in RAW and JPG under the same conditions and you can examine what happens.
JPG is mostly darkening the image resulting in less highlight blinkies and more lowlight blinkies. Until now I didn't get more highlight blinkies.
Another way to check the blinking issue is to compare the on-camera blinky-screen with the blinky-screen in the RAW-converter. But as far as I know that's only possible with Nikon and Capture.

George

[ajohnw]

Your seem to be confusing 2 different things Geogre, dynamic range differences between raw and visual images/jpg's and dynamic range in a raw file against transfer characteristics used in cameras and PP packages to generate a visual image.

If a subject uses up all of the dynamic range available in the raw file it could be mapped as described directly to a jpg but it wouldn't look very good. If the dynamic range in the subject just falls short of generating blinkies or even well short the mappings that are used will give a decent image directly in an ideal world. I'm not mentioning problems around the blinky point at the dark end. Separate aspect.

In some ways there is little point in the post that shows different ;levels of clipping. For one is it a channel, all channels, 2 channels or what? The main reasons for the difference are what happens to the raw data which is loosely speaking as follows

Raw file -> Camera Colour Profile -> Colour Temperature Correction -> Mapping to the visual image / jpg etc

If Nikon camera and Adobe are compared only the raw file and colour temperature correction are the same with some doubts about the latter being exactly the same. Actually it's the same with Canon as well and probably all cameras as Adobe do what they feel suits the camera. The manufacturers do their own thing. Packages do their own thing too which is why they produce different initial results but in principle all could be used to generate an image which to all visual aspects are identical. Start measuring pixel values and there are likely to be differences as each package is very likely to use different methods of getting there.

If you find a package changes the image when you EXPORT to jpg and reload something is decidedly wrong. The whole point of a PP package is to actually show what you will get when the image is viewed on anything in principle but monitor calibration or lack of it etc can mean that different people will see different things. A common reason for apparent changes in brightness in images is the colour of the background also the work that some one has just done on an image. That tends to concentrate peoples attention to what ever they are adjusting - not the entire image.

John
-