JPEG vs RAW dynamic and color range

[John King]

G'day again Matija

No. A wider gamut is capable of rendering more colours than a narrower gamut. This is fundamental to the science of colour spaces.

Both a wide gamut and a standard gamut monitor can render the exact same amount of colours - 2^24, or 16.777.216.

Monitor gamut is not a measure of the amount of colours, only of their positioning and intensity.

The output from your computer is 8 bits per channel. It doesn't matter if your monitor has 4398-bit processing or a 2387-bit panel (those numbers don't mean what you think they mean; they're used for dithering and 3D LUT transforms, to mitigate all that posterization and banding that would otherwise be inevitable).

At the end of the day, there are 2^24 colours on a CRT, 2^24 on an sRGB monitor and 2^24 on a wide gamut monitor, simply because that is the only thing the computer knows how to send to the monitor - hence my statement that wide gamut is inherently flawed under the current display path, which is 8-bit-per-channel. Once that changes, everything will be vastly different. If it changes at all, because there's more and more shifting back to sRGB from monitor manufacturers.

Well now. The video cards I use have 32 bit output (theoretically ... ) and a 64 bit data path on the input side (in/out x 32 bits each), IIRC.
I wonder why they design them that way?
The monitor uses DVI input. How many bits is that?

I have not seen any video card with less than 16 bit addressing for at least ten or twelve years, although I do keep a couple of spares (of both 8 and 16 bit cards, with drivers) for clients who might need them, or would have to trash an otherwise perfectly suitable computer (for them ... ).

BTW, my monitor is switchable from aRGB to sRGB modes, and guess what? The difference in colour is quite dramatic ...

--
Regards, john from Melbourne, Australia.
(see profile for current gear)
Please do not embed images from my web site without prior permission
I consider this to be a breach of my copyright.
-- -- --

The Camera doth not make the Man (or Woman) ...
Perhaps being kind to cats, dogs & children does ...

Gallery: http://canopuscomputing.com.au/gallery2/main.php



Bird Control Officers on active service.

Member of UK (and abroad) Photo Safari Group

 

[John King]

Gidday Bill

After reading your responses and the replies of others from this point down, methinks that you've drunk and swallowed the ARGB Kool Aid.

I will let that one go through to the keeper ...

There appears to be some "it's there but you can't see it" defense but, well....

See my response to Matija re my monitor being switchable from aRGB to sRGB. The difference is very noticeable; i.e. one can see it plainly on the screen.

--
Regards, john from Melbourne, Australia.
(see profile for current gear)
Please do not embed images from my web site without prior permission
I consider this to be a breach of my copyright.
-- -- --

The Camera doth not make the Man (or Woman) ...
Perhaps being kind to cats, dogs & children does ...

Gallery: http://canopuscomputing.com.au/gallery2/main.php



Bird Control Officers on active service.

Member of UK (and abroad) Photo Safari Group

 

[MatijaK]

Well now. The video cards I use have 32 bit output (theoretically ... ) and a 64 bit data path on the input side (in/out x 32 bits each), IIRC.

Nope. The output is 8 bits per channel, making for a total of 24 bits, or 16.777.216 colours.

The last 8 bits are internally represented as "alpha" (transparency) inside the graphics card, for a very simple reason - working with a 32-bit quad is many times faster than working with a double and a single.

The monitor uses DVI input. How many bits is that?

8 bits per channel, which is what your graphics card sends out to the monitor.

BTW, my monitor is switchable from aRGB to sRGB modes, and guess what? The difference in colour is quite dramatic ...

Nobody said otherwise

Go back a little, to this statement:

R = 156, G = 64, B = 41 (PPRGB) vs R = 226, G = 0, B = 37 (sRGB)

To get a certain intensity of red, you had 156 shades of actual red in wider gamut, compared to 226 in sRGB, which is a 44% reduction in dynamic range. (We'll ignore G&B this time because they're just there to taint the colour for the monitor's gamut.)

 

[Bootstrap]

I respect your opinion and certainly your right to state it but you'll not convince me.

As with so many other things photographic, much ado about nothing. If the end user cannot see the results, why bother?

Bill Turner

 

[rovingtim]

aRGB has the biggest colour gamut, which means it can display colours that neither sRGB or screens can display.

sRGB is has a smaller colour gamut, but it is still capable of displaying colours that most prints (CMYK) are unable to reproduce. However, it is only slightly larger/different than prints so it is best for editing for print.

Print has the smallest colour gamut. Trying to reproduce colours that are not within the print's colour gamut can do strange things to the final print colour. That is why most people avoid aRGB unless they understand which colours are available for print and most screens.

Google colour gamut for more.

 

[luisflorit]

I think Olympus is able to extract more from the camera than industry standard ACR.

No doubt raw contains more data, but the question is which software can read all or most of it correctly.

Could you post the orf file somewhere? I can give it a try with RawTherapee.

L.

--
My gallery: http://luis.impa.br/photos



Oly E5/E3 + 12-60 + 50-200 + EC14 + EC20 + FL50R
Pany FZ50 + Oly FL50 + TCON17 + Raynox 150 & 250

 

[tinternaut]

I mostly shoot JPEG but sometimes dabble in raw. Here's my experience using Aperture with files from my E-30 and E-PL1:

• Firstly, if the JPEG has minor blown highlights, I find they are easily recoverable from a shot taken at ISO100
• At ISO100, Aperture's default rendering is a little on the bright side anyway so a boost adjustment, -0.3 E/V on the exposure slider and a slight movement of the recovery slider normally does the trick
• At ISO200, it gets a bit more complicated
• At ISO200, there may be some recoverable highlight data but it can sometimes be tricky to recover (software with layers and masks would probably make it easier)
• For the ISO200 shots, I often find it best to recover everything I can and then approach getting the picture to pop using a tone curve adjustment to both increase brightness and contrast
• Whereas at ISO100, the post processing regime is the same as per my JPEGs, starting with a levels adjustment.

This is very different to using the latest version of ACR/PSE which I find much easier to use on ISO200 raws. Aperture OTOH is transparent with my ISO100 raws (one of the reason I'll always take an extra couple of ISO100 shots where I think what I'm shooting will be a keeper.

Why the difference between ISO100 and ISO200? Well the various raw converters try to reflect what Olympus does in terms of the exposure and JPEG processing to expand the highlights in JPEGs taken at ISO200 and some achieve this better (at least in terms of ease of use) than others.

--
Regards
J

Follow me on Twitter: http://twitter.com/jasonhindleuk
Blog: http://jasonhindle.wordpress.com



Photos: http://www.flickr.com/photos/jason_hindle

Gear in profile

 

[msusic]

Here's a test for you who say RAW is always superior to JPEG.

I took this shot of a friend today, low sun caused the image to look red(ish), certainly similar to the conditions I shot it in.

JPEG has virtually no clipping, there is a tiny bit which is easily fixed by small decrease of vibrance/saturation by few points.

On the other hand, RAW is severely clipped and I have tried desaturating yellows, reds and oranges while applying recovery only to end up with horrible results and the clipping remains while the image turns almost void of red/yellow/orange color, certainly nothing like it looked when I took it.

Here's the raw:
https://rapidshare.com/files/3539429704/P5219688.ORF

And full sized OOC JPEG:
http://www.pohrani.com/f/2y/cy/2IFUUOZa/p5219688.jpg

I'd be interested in seeing RAW conversion without clipping using ACR.
--
Cheers,
Marin

 

[Andy Hewitt]

Here's a test for you who say RAW is always superior to JPEG.

I took this shot of a friend today, low sun caused the image to look red(ish), certainly similar to the conditions I shot it in.

JPEG has virtually no clipping, there is a tiny bit which is easily fixed by small decrease of vibrance/saturation by few points.

On the other hand, RAW is severely clipped and I have tried desaturating yellows, reds and oranges while applying recovery only to end up with horrible results and the clipping remains while the image turns almost void of red/yellow/orange color, certainly nothing like it looked when I took it.

Oh fer goodness sake man.

The JPEG is derived from the Raw data that is produced by the camera (which also ends up as a Raw file if that's what you save as) - the JPEG CAN NOT contain more data than the Raw file.

What you're seeing is the difference in how the data is being processed after the Raw data is created. Note that Raw files are NOT an image file, they are Raw data. They have to be converted/interpreted into something that can be viewed as an image.

--
Andy Hewitt

 

[msusic]

I'm well aware of that
Just asking people to try to make it better with ACR or other RAW converters.

Yes, RAW data contains more than a JPEG, but it's also important the way it's processed. If there are no perfect raw converters around and if they're worse than JPEG in extracting tonal and dynamic range, I can claim that JPEG is better because other alternatives are worse.

Adobe is mainly interested in Canikon camp and I wouldn't be surprised that they can't extract as much from the sensor in terms of color and dynamic range as the JPEG engine (wrote by Olympus can).

Here's a test for you who say RAW is always superior to JPEG.

I took this shot of a friend today, low sun caused the image to look red(ish), certainly similar to the conditions I shot it in.

JPEG has virtually no clipping, there is a tiny bit which is easily fixed by small decrease of vibrance/saturation by few points.

On the other hand, RAW is severely clipped and I have tried desaturating yellows, reds and oranges while applying recovery only to end up with horrible results and the clipping remains while the image turns almost void of red/yellow/orange color, certainly nothing like it looked when I took it.

Oh fer goodness sake man.

The JPEG is derived from the Raw data that is produced by the camera (which also ends up as a Raw file if that's what you save as) - the JPEG CAN NOT contain more data than the Raw file.

What you're seeing is the difference in how the data is being processed after the Raw data is created. Note that Raw files are NOT an image file, they are Raw data. They have to be converted/interpreted into something that can be viewed as an image.

--
Andy Hewitt

--
Cheers,
Marin

 

[Andy Hewitt]

I'd be interested in seeing RAW conversion without clipping using ACR.

Ok, I downloaded those two. Your JPG had clipped highlights, and was over saturated, and, IMHO, had too much warmth on the white balance. You could see where detail had been lost on the girl's hair and dress, and the hair was tinted, again IMHO.



Working on the Raw file, it was easy to recover the clipped highlights, and adjust exposure and levels to bring everything back.

Here's my version using Aperture, although I think I overdid recovering the shadows. I think the hair and skin tones look much more natural though.



--
Andy Hewitt

 

[msusic]

The entire image looks washed out and gray-ish now.
I'd like to see image with tones close to the JPEG, but without clipping.

BTW, direct low Sun casts very warm tones on the faces.
--
Cheers,
Marin

 

[Andy Hewitt]

The entire image looks washed out and gray-ish now.
I'd like to see image with tones close to the JPEG, but without clipping.

Hmmm, how is your monitor calibrated? Mine is setup with a Pantone Huey, and the picture looks OK, yours looks over-saturated to me. If you've got a different calibration to me, which you almost certainly will have (or none at all), then we will see this differently.

BTW, direct low Sun casts very warm tones on the faces.

Yeah, I know that, but your original appears a bit overdone. Aperture adjusted (I used the inside of the tent roof for a grey point) it to about 4080deg, which is about 1000deg lower than the original.

What you actually want is software that'll emulate the JPEG engine that's in the camera, yes? Have you tried Olympus Viewer?

--
Andy Hewitt

 

[John Mason]

I typically don't like the out of the camera Jpegs compared to raw processing. Depending on the subject mater I use Capture One most often and LR sometimes.

I use a wide gamut monitor and printer. I use the out of the camera Jpegs as proofs to select which ones to work with in Raw.

I'm sure I could work with the Jpegs as a starting point, but the workflow in Capture One is so efficient (and only about a sec per pic to develop) there is no advantage time wise to using the Jpegs as a starting point.

On most pictures there are no gamut challenges such that sRGB doesn't compromise. And if you've got good software, the interpolation of out of gamut material back into sRGB is such that even with out of gamut material it's doubtful the end user of the picture would see much difference.

I find most out of gamut stuff where I prefer the wide gamut setup is in long exposure night photography and in flowers. Normal portraits there isn't going to be much in the photo to make any difference in my experience.

One such picture that made a large difference was a rainbow shot over Athens. I'll pop up the sRGB version here:





The real image when proofing with qImage at the full gamut of my monitor or on the actual print shows more colors and looks more 'real'. (easy to see when side by side).

But most shots it isn't going to matter or be visible in a final print. (and those that would see a difference even if there is a difference will be an even smaller population sample)

--
John Mason - Lafayette, IN

http://www.fototime.com/inv/407B931C53A9D9D

 

[Mescalamba]

Your issue with RAW vs JPEG isnt bad RAW, but RAW developing program. ACR is far from being best (apparently especially for Olympus). LightRoom is bit better in recovering details, but from my experiences other RAW developing programs are sometimes better than ACR/LR.

For example, I used for long time LR to develop my m4/3 Pany RAWs with somewhat mixed results (usually not so good colors, bit washed out, low-contrast). Then I switched to Capture One and voila its exactly as I shot it, or even better.

Exotic things like DCraw can yield sometimes very interesting results (and in my case usually very good). Theres RAW Therapee SW based on DCraw too, if you want something more user friendly.

And I kind dunno why you dont use Olympus Master, its supposed to be best converter for Olympus?

Otherwise, if you shoot RAW+JPEG and JPEG is correct and RAW not, its cause you are not using proper developing, cause you can bet that those JPEG data are inside that RAW.

 

[John King]

Gidday Matija

Well now. The video cards I use have 32 bit output (theoretically ... ) and a 64 bit data path on the input side (in/out x 32 bits each), IIRC.

Nope. The output is 8 bits per channel, making for a total of 24 bits, or 16.777.216 colours.

Check the data sheets ...

The last 8 bits are internally represented as "alpha" (transparency) inside the graphics card, for a very simple reason - working with a 32-bit quad is many times faster than working with a double and a single.

The monitor uses DVI input. How many bits is that?

8 bits per channel, which is what your graphics card sends out to the monitor.

Source?

BTW, my monitor is switchable from aRGB to sRGB modes, and guess what? The difference in colour is quite dramatic ...

Nobody said otherwise

Go back a little, to this statement:

R = 156, G = 64, B = 41 (PPRGB) vs R = 226, G = 0, B = 37 (sRGB)

To get a certain intensity of red, you had 156 shades of actual red in wider gamut, compared to 226 in sRGB, which is a 44% reduction in dynamic range. (We'll ignore G&B this time because they're just there to taint the colour for the monitor's gamut.)

No. The sRGB image has far too much red, and far too little green. The colour balance (as shown by the channel numbers) is patently wrong in the sRGB image.

The PPRGB image is as near as it can be to the actual colour of the car, both on screen and in print. The sRGB image is nothing like the real colour. So it appears to me that you are arguing the matter backwards.

Have you downloaded the two images I have uploaded and posted and done the measurements in a properly colour managed environment?

I have just taken precise measurements of the two images at the locations noted (as close as I could get to identical). The actual figures are:

R = 153, G = 61, B = 36 (PPRGB at X = 2.80, Y = 3.68)
vs
R = 221, G = 2, B = 32 (sRGB at X = 2.79, Y = 3.68)

It is the over-abundance of red and the shortfall of green (these proportions are reflected across the entire image) that causes the colour shift to orange, and the apparent over-exposure in the print. Both of these effects are also apparent on my monitor.

You can explain this away anyway you like; but I am trying to explain it, not explain it away ...

BTW, print technologies have come a long way in the seven years since Blatner & Fraser's book was published in 2004 (the edition I have). Even recently, Seiko Epson out-sourced the development of the new colour LUT for the Rx880 series to accommodate the wider gamut that these printers are capable of. One compelling reason for my buying the R3880 was the much larger colour gamut over previous models and other brands.

My observations are severely practical; not some theoretical musings. I am trying to explain the clear differences I see on my monitor and in print between these different colour spaces - differences that are also clearly apparent to other people to whom I have shown these images and the test prints resulting. I am not trying to be argumentative, merely sharing my personal observations, and seeking further elucidation.

--
Regards, john from Melbourne, Australia.
(see profile for current gear)
Please do not embed images from my web site without prior permission
I consider this to be a breach of my copyright.
-- -- --

The Camera doth not make the Man (or Woman) ...
Perhaps being kind to cats, dogs & children does ...

Gallery: http://canopuscomputing.com.au/gallery2/main.php



Bird Control Officers on active service.

Member of UK (and abroad) Photo Safari Group

 

[luisflorit]

Sorry, but your JPG is heavily clipped. The fact that you see little in the 255 range does not mean it is not clipped. It has a lousy recovering done. The girl face looks quite bad in your shot, IMO.

Actually, I think your picture is the perfect example of a situation where RAW can save the picture...

When I open your image in RT and select NEUTRAL (that if, if I do NOTHING), I got a very good almost non-clipped shot. Clicking in "recover highlights", and everything is recovered (except of course for the sky). A small bump in brightness, and I end up with a very natural looking shot. Of course, you can work a lot over it if you want, saturating reads or bumping contrast, for example.

Take a look (everything only resized). Your shot :



RT in neutral, with nothing applied:



Recovering highlights:



Bumping brightness:



Your shot again:



Try RawTherapee. It's free.

Cheers,
L.

--
My gallery: http://luis.impa.br/photos



Oly E5/E3 + 12-60 + 50-200 + EC14 + EC20 + FL50R
Pany FZ50 + Oly FL50 + TCON17 + Raynox 150 & 250

 

[Raist3d]

it's not a simple 12 bit to 8 bits. This is a very common misconception about JPEG.

Of course. But it is a still a lossy conversion.

Yes, but it's not anywhere near as lossy as some people make it seem. In fact, as I have said many times, the images that people show on the web "coming from RAW" are almost all the time JPEGs, not even a lossless format like PNG.

Data is encoded in weight, in areas. That's why JPEGS look as good as they do. In fact, everything we share and when people go "ooh, ahh look at this raw output" is 99% of the time on the web a JPEG.

If you look at a ProPhotoRGB image on the web, it looks abysmal. e.g.

[del]

Not sure at all what was your point here showing a pro photo rgb, and sRGB image. A pro Photo RGB image would look abysmal on the web if you don't have a browser that color profile converts on the fly to what the monitor you are using uses (usually the sRGB).

As for sRGB- aRGB can have some advantages when printing that vary wildly by subject and by the quality of print. But personally I don't find it worth the workflow hassle. To really see it on a monitor you need to get an adobe-gamut aRGB monitor which are expensive and if you are going to share on the web now you are adjusting colors to something that most of your audience will see it differently and you don't even know it (unless you compare in sRGB / aRGB browser profile aware etc in a normal monitor).

BTW, I do use a monitor with an aRGB gamut (24 inch, it wasn't terribly expensive - under Oz $1,000); and print on an Epson R3880. I do notice a big difference between colour spaces.

I didn't say you necessarily would or not. My point was, you will need an Adobe rgb if you are doing this to color correct properly unless you like printing a lot.

Most people who look at my images do so on my monitor, or my A4~A2 prints ...

I also convert almost everything I upload to my web site to sRGB, but I do not use it otherwise.

Not sure again, the relevance of this.

It hasn't been for me the workflow hassle but if it works for you great.

What hassle? There isn't any. None. Zip. Nada. Zero.
Also BTW, one does not "adjust colours"; one converts to a colour space …

I am not talking about the conversion. I am talking that if you are post processing/color correcting/doing contrast color adjustment to an image with aRGB data, but you only have an sRGB monitor, you will have the hassle now that what you print is not what you were seeing on your monitor (assuming it's color calibrated).

That is the hassle.

Now if this isn't an issue for your needs, well then, great for you.

--
Regards, john from Melbourne, Australia.
(see profile for current gear)
Please do not embed images from my web site without prior permission
I consider this to be a breach of my copyright.

--

Raist3d/Ricardo (Photographer, software dev.)- "You are taking life too seriously if it bugs you in some way that a guy quotes himself in the .sig quote" - Ricardo

 

[rovingtim]

I did two because some prefer images to be white balanced while others prefer the orange light. I prefer orange light.

Here's yours after moving two sliders (clarity and vibrance). The clarity slider I used to personal taste ... I could turn it off if you want to see another. I also used full highlight recovery on this one -- which didn't recover anymore than I'm used to with the E3, by the way. This is not white balanced ...





And here's one that was white balanced. This one took one click and I was playing with the sharpening (which I didn't mean to leave on ... but it is sharpened more than default. Oh well.) I'm afraid I slid the clarity slider down on this one as well.





And here's yours.



Thoughts? I can adjust it in other ways if you prefer. These conversions took almost a minute.

 

[John King]

it's not a simple 12 bit to 8 bits. This is a very common misconception about JPEG.

Of course. But it is a still a lossy conversion.

Yes, but it's not anywhere near as lossy as some people make it seem.

Considering that a 10 MP, 8 bit file "should" be 10 MP x 3 bytes per pixel, then losslessly compressed at ??. I will leave you to calculate the data loss ...

In fact, as I have said many times, the images that people show on the web "coming from RAW" are almost all the time JPEGs, not even a lossless format like PNG.

This has just about zero relevance, as both size and colour space are almost always very limited for display anywhere on the web.

Data is encoded in weight, in areas. That's why JPEGS look as good as they do. In fact, everything we share and when people go "ooh, ahh look at this raw output" is 99% of the time on the web a JPEG.

If you look at a ProPhotoRGB image on the web, it looks abysmal. e.g.

[del]

Not sure at all what was your point here showing a pro photo rgb, and sRGB image. A pro Photo RGB image would look abysmal on the web if you don't have a browser that color profile converts on the fly to what the monitor you are using uses (usually the sRGB).

If you read what I wrote, rather than just reacting, you would have downloaded the image/s and examined them in a properly colour managed environment. It also demonstrated that one cannot currently display web images using anything BUT sRGB. This will change as surely as inkjet printers went from being monochrome to colour; as did monitors; etc ...

As for sRGB- aRGB can have some advantages when printing that vary wildly by subject and by the quality of print. But personally I don't find it worth the workflow hassle.

Does this change anything I have said? No.
You cannot be bothered. I can. So can others.
I 'bother' because I can see significant improvement in print and on-screen.

It really isn't any effort at all. You make out that it is a great effort. It isn't.

To really see it on a monitor you need to get an adobe-gamut aRGB monitor which are expensive

I have just told you that even in Oz, the price of these monitors has plummeted in the last 12 months - $2,600 -> $699. 20 years ago, a Phaser colour printer that would print photo quality A3 cost slightly less than $100,000. My Epson R3880 cost less than $2,000; will print to A2 size; and at superior quality (I have a Phaser print for comparison ... ). Things change. We should try to change with them ...

and if you are going to share on the web now you are adjusting colors to something that most of your audience will see it differently and you don't even know it (unless you compare in sRGB / aRGB browser profile aware etc in a normal monitor).

You appear to misunderstand this. One edits in a wide gamut colour space, then converts the colour space using (say ... ) a Relative Colorimetric rendering, which places those wide gamut colour numbers appropriately within that tiny little sRGB colour space ...

BTW, I do use a monitor with an aRGB gamut (24 inch, it wasn't terribly expensive - under Oz $1,000); and print on an Epson R3880. I do notice a big difference between colour spaces.

I didn't say you necessarily would or not. My point was, you will need an Adobe rgb if you are doing this to color correct properly unless you like printing a lot.

I have done this very successfully for some years using a CRT that was supposed to only have an sRGB gamut. Never had a print problem from this cause.

Most people who look at my images do so on my monitor, or my A4~A2 prints ...

I also convert almost everything I upload to my web site to sRGB, but I do not use it otherwise.

Not sure again, the relevance of this.

I do not know why you cannot see what I am saying. Selective colour blindness, perhaps ...

It hasn't been for me the workflow hassle but if it works for you great.

What hassle? There isn't any. None. Zip. Nada. Zero.
Also BTW, one does not "adjust colours"; one converts to a colour space …

I am not talking about the conversion. I am talking that if you are post processing/color correcting/doing contrast color adjustment to an image with aRGB data, but you only have an sRGB monitor, you will have the hassle now that what you print is not what you were seeing on your monitor (assuming it's color calibrated).

Yes, it is. I work in a fully colour managed environment.

That is the hassle.

Yes, it would be. But I have never had to do this ... Perhaps working in a wide gamut, 16 bit colour space and RAW has its benefits?

Now if this isn't an issue for your needs, well then, great for you.

It never has been, thusly far.

--
Regards, john from Melbourne, Australia.
(see profile for current gear)
Please do not embed images from my web site without prior permission
I consider this to be a breach of my copyright.
-- -- --

The Camera doth not make the Man (or Woman) ...
Perhaps being kind to cats, dogs & children does ...

Gallery: http://canopuscomputing.com.au/gallery2/main.php



Bird Control Officers on active service.

Member of UK (and abroad) Photo Safari Group