Oh no, not Adobe RGB v sRGB again, groan

[crisscross]

sorry yes, cos eventually I was persuaded by deluge of opinion here and elsewhere that one should always convert to sRGB for the web and that it made virtually no difference on normal good monitors....
....until I was startled by horrid effect of pressing the convert button on this image
Maybe it selectively relies on all the yellow-green and russet ranges that get cut out in sRGB, or other ideas welcome
aRGB first

and sRGB

PS although this is done in NX2, I checked it out in CS4 with same result

[nanook]

Oh my. Deep colours obviously better on the top one. The yellows are intense in both renderings, but without the deepness of the trees or the grass, it is only yellow dominating. Difficult image, high contrasts. As JPEG is 8-bit, I didn't think any conversion would be this visible.

I have run a few test from Lightroom 3, and it is obvious that sRGB in some cases give less than a desired result. My suggestion, based on the provoking comparison you here put forth (sorry, read a Dickens just recently), would be to render any file in as wide a gamut as possible when intended for screen. It does not hurt in my opinion. Prophoto RGB encompasses the entire sRGB and then a lot more.

No colourspace will guarantee you desired results on the web, but if I had watched these two photos at work, with a monitor that cannot ever render sRGB, I would not see any difference between the two. Then again, why should you care? The wider gamut on the top will however ensure that if I have a decent monitor (and I'm not even talking expensive or exclusive) - you would give me a better viewing experience with the top one.

However, as the white points are different between the colour spaces as well, another effect of conversion will be contrast and brightness. The bottom one obviously has less contrast and an altogether lighter look. I will not even start to pretend the conversion algorithm between ARGB and sRGB, but it is not straightforward.

The upper photo loses a bit of detail at this resolution in the dark shadows in my opinion and the greens are a tad yellow still - but nothing compared to the bottom one. The bottom one loses detail in the highs, but retains more detail in the shadows. What draws the eye in the top one is the foilage, while in the bottom one it is the sheeps. erm... It does look a bit burned out in the highlights, losing the perception of contrast.

When rendering for printers however, you must cater for one and only one printer and then colour profiling is very very important. I don't know where to throw my parents HP C5180, but it will not be a pleasant experience for the horrid piece of plastic.

You would have great pleasure in reading the details on the JPEG article in wikipedia. Different formats explained. Only newer jpgs can retain EXIF-info, ICC profiles. As they can contain ICC-profiles however, doing the image AdobeRGB will ensure a better conversion to the viewers ICC than from an sRGB - as sRGB cannot upsample its colourspace.

Firefox 3.5 and above support ICC profiles. So including a wider colourspace than sRGB can in fact have beneficial effects.

Sorry if I ramble. Interesting comparing the images.

[kevinf]

Change the rendering intent to saturation when converting to srgb.

[crisscross]

Thank you for your detailed commentary Espen, which confirms my understanding of the technicalities and that what I see is what I see. The printing problem is different as I only print from the original aRGB, and although on my cheap iX4000 printer the SP4 profile does not actually match the best Canon papers, it is adequate using HP (!)

Change the rendering intent to saturation when converting to srgb.

Kevin - it makes no difference in NX2 (which I will have to take up with Nikon). In CS4 it does appear to work using 'working profile sRGB IEC6..etc' (as opposed to just sRGB IEetc), then 'save for web' in which dialogue also re-embedding the sRGB profile. Odd, so I also tried doing it in PSE4 and there also you get the dull version.

So it seems to be necessary to use a 16bit tif as an intermediary and using quite a complex sequence of operations

But more importantly it raises a further spectre of when is sRGB really sRGB and when is it a bastard....EXIF tool on the CS4 generated jpg clearly identifies the colour space as the genuine sRGB IEC61966-2.1 copyright HP etc. The Nikon generated one is just uncalibrated. What is the point of an international standard that is copyrighted so no-one else can use it???

So how does it come out at the users end, for which here is the version processed in CS4 with all the checkboxes co-ordinated:

[crisscross]

PS for any NX2 users reading - Nikon sorted the problem: it is essential that the adjust>colour profile>srgb is done as the very last step. If as often happens one sees something else, delete the original step and re-add at the end