Pages: [1]   Go Down

Author Topic: The Dreaded JPEG export saturation question again  (Read 8600 times)

joedecker

  • Full Member
  • ***
  • Offline Offline
  • Posts: 142
    • http://www.rockslidephoto.com
The Dreaded JPEG export saturation question again
« on: July 13, 2009, 09:05:16 pm »

I just ran a set of experiments, exporting the same (original LAB PSD file) via both LR and PS. For PS I used "save for web" after a convert to mode using the perceptual rendering intent, separately with and without an embedded profile. For LR I just used export testing all three color space options.

On color-managed browsers, on my sample, the best results were those that either were produced by LR, or those produced by PS with an embedded profile. That these two seemed equivalent should be no surprise, I surmise (and assume as we go on here) that LR embeds color profiles.  For images with embedded profiles, on color-managed browsers, the results seemed very similar, again, I expected this result, most monitors can't display much in the way of colors outside of sRGB, as I understand it.

On non-color managed browsers the best (err, most saturated) results were those for which there was an embedded profile (LR or PS) and for which the smallest [sic] color space had been specified.   No difference was observed between those with and those without color profiles, which is no surprise, that's what we mean by "non-color managed."  If you think about it, that we'd get the best color saturation from smaller spaces makes sense. A modest red might be "75%" red in RGB, a modest red might be "30% red" in a larger color space. Since non-color-manged browsers just stick "75" or "30" to the graphics card without context, of course the 75 is going to look more ... red.

On non-color managed browsers, it seemed like I was seeing a little more washout near the lightest (most luminous) reds on files prepared by LR relative to those prepared by PS. I am unable to account for this washout directly, since I had read suggestions that LR uses the perceptual rendering intent, but perhaps that's wrong or perhaps the input color space to perceptual conversion matters more than I expect.  I did see hue differences between LR and PS results on non-color managed browsers, particularly around 25% L between blue and red, as well as color shifts in the fully-saturated greens (the PS results seemed to peg at a warmer green).

LAB images in PS directly exported appeared perhaps identical to sRGB exports without embedded profiles; PS "Save for web..." greys out the profile embedding option on LAB images.

When this subject has been discussed on the web before it's way too much fun to simply suggest "get a real browser", but that's not a very practical solution for folks who are preparing images for the world wide web.  While one could throw up ones hands and simply suggest "there's nothing to be done at the mess the web is in", and there is some truth to that, it makes no sense to me to entirely ignore how JPEGs will appear on non-color managed browsers.

So, this pedantry aside,, this suggests, at least to me, the following recommendations for highest saturation results for general web usage:

1. Use Photoshop, not Lightroom, for JPEG export.   I don't know why this gives better results, but it appears to.  This choice benefits (and I don't know why) non-color-managed browsers.

2. Convert to sRGB using "Convert to Profile". Choose the rendering intent that looks best to you.  The choice of sRGB will benefit the vast majority of WWW users on non-color-managed browsers.  (If you must use LR in step 1, this still applies.)

3. Use "Save for web..." and embed the profile. This last step will slightly increase the size of your file but make your image notably peppier for about 10% of your users (Safari users and the few of us using Firefox with CM turned on.)

(Obviously, the subjectively best option for a specific image may be different.)

Testing details:

I worked myself from an image I produced in PS in LAB using a spectral gradient left-to-right, then overrode the L channel in LAB with a black-to-white gradient vertically, and I think I've explained above the rest of the options and procedures used.  I used Safari on a Windows XP to test on color-managed browsers, and Google Chrome and IE8 to test on non-color managed browsers.

I'm entirely open to the idea that I've screwed up this test.  Comments and criticisms welcome.

Joe
Logged
Joe Decker
Rock Slide Photography [url=h

jbrembat

  • Full Member
  • ***
  • Offline Offline
  • Posts: 177
The Dreaded JPEG export saturation question again
« Reply #1 on: July 14, 2009, 03:30:08 pm »

I'm not interested on Adobe products, but the following sentence is incorrect:

Quote
Convert to sRGB using "Convert to Profile". Choose the rendering intent that looks best to you
sRGB is a matrix profile, so the rendering intent used is relative.

To have perceptual, you must use sRGB_V4 profile.

Jacopo
Logged

joedecker

  • Full Member
  • ***
  • Offline Offline
  • Posts: 142
    • http://www.rockslidephoto.com
The Dreaded JPEG export saturation question again
« Reply #2 on: July 15, 2009, 12:39:51 pm »

Quote from: jbrembat
I'm not interested on Adobe products, but the following sentence is incorrect:


sRGB is a matrix profile, so the rendering intent used is relative.

To have perceptual, you must use sRGB_V4 profile.

Jacopo

You're right, of course.  Thanks.
Logged
Joe Decker
Rock Slide Photography [url=h

Tyler Mallory

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 62
    • http://www.tylermallory.com
The Dreaded JPEG export saturation question again
« Reply #3 on: July 16, 2009, 08:38:04 pm »

Just a thought:
I've noticed that, when saving to a JPEG, using the format option of "Baseline Optimized" seems to result in slightly better color rendering. Is it a case that the "Save to Web" function is utilizing that format, but LR is not?
Pages: [1]   Go Up