Taming (profiling) the wild iPad

[rasworth]

I created a profile I believe represents the colorspace of at least one (mine) first generation iPad.  I'm not going into details as to the dervation method - it involves displaying and measuring a large group (99) of color patches on the iPad, and then using the spectral measurements as input to a profile generating program.  Suffice to say any good hacker should be able to figure out how to do such, although there are some traps to avoid (make sure 32 bit color comes across, turn off auto brightness on the iPad, etc.).

As most iPad owners know the display is already superb by laptop/tablet standards, using an IPS LCD panel that's well controlled.  However, the gamut is smaller than sRGB, as evidenced by the attached gamut plots (Yxy and xy chromaticity, wireframe is sRGB).

I used the profile to convert several images for transfer to the iPad to test its usefulness.  Probably more significant than the hue shifts is the tonal response; like most LCD monitors the iPad display tends to squeeze the shadows, and conversion using the profile will bring out detail in darker areas, such as hair in portraits.  The third attachment, neutral curves, clearly shows the "bend" required on the RGB channels necessary for tonal correction.

And finally a before and after of a colorchecker image, upper left of each patch is straight sRGB values and lower right is converted via the iPad profile.  There are definite hue shifts, although again the tonal differences are probably more significant.  This image has no embedded profile, for those of you with iPads please save it off and load it on your iPad for comparison.  Also remember some clipping is occurring, the cc cyan patch (third row, far right column) is out of gamut in sRGB (red channel clipped to 0), and the yellow patch is additionally out of gamut (blue channel clipped to 0) for the iPad space.

Richard Southworth

[rasworth]

I added a real world example, sRGB carried straight across with no change on the left, and the same image converted thru the iPad profile on the right.  The most apparent difference is in the darker areas, in terms of retaining detail.  Once can also see some saturation and hue shifts, although they are more subtle.

Again, this image contains no embedded profile, and should be transferred across to an iPad for "true" viewing.

Richard Southworth

[francois]

On my monitor, the difference is obvious and not only in the dark regions. The reds are also more saturated - or less muted.

[rasworth]

The image was created to be viewed on an iPad, and doing such on most monitors will tend to exaggerate the differences, although still providing a reasonable relative comparison.  I'm hoping that some forum members will download and transfer to their iPads, and give me some feedback as to whether or not its worth the effort.

Apple sold 15 million iPads last year, and is projected to sell 40+ million this year.  I'm convinced that more and more people are going to want the brunt of their good images (paid for portraiture, model portfolios, etc.) "printed" to iPads, and therefore we should learn how to best manage.

I have a client that generates most of his revenue from high school students pictures.  He has already seen a significant decline in demand for wallet size prints, rather his customers want small digital renderings for their smart phones.  I believe that in a short time paid for prints will be relatively small volume large framed wall type.

Richard Southworth

[fike]

So did you change the embedded profile applied from sRGB (or Adobe RGB or Pro Photo RGB) to something you created like, for example, Richard's iPad RGB?

[rasworth]

Per the initial post, I created a profile that represented the iPad colorspace.  The test image started out in sRGB with an embedded profile.  I made two copies of the image - I assigned the iPad profile to the left hand image (did not change the actual pixel values) and did a Convert to Profile on the right hand image (did change the pixel values).  I then stitched the two images together and saved without an embedded profile.

When transferred to an iPad, the left hand side is just as if the sRGB image was transferred with no modification, and the right hand side is rendered for the iPad colorspace, thereby providing a side by side comparison of without and with iPad profiling.

Richard Southworth

[kimballistic]

On the ipad 2 (I know, not the device you profiled :-), the differences in the photo of the woman are minimal.  You really have to search for them.  The image made with the ipad profile has slightly brighter greens/shadows (the leaves) and more saturated oranges/reds (the flower).  No difference in skin tone or hair.  The differences in the color picker chart are much more obvious of course.

To be honest, the resolution and sharpness of the file is pretty basic.  Why not demonstrate with a full portrait instead of a small crop of one?  Just my $0.02.

[rasworth]

Thanks for the feedback, what you describe is pretty much what I see on an iPad 1.  I cropped down the picture because (1) I didn't want to reveal the face (2) just trying to concentrate on the shadow detail.

I'll look for something else to compare; however, as I stated before the display is already good, the profile conversion only makes subtle changes.

Richard Southworth

[rasworth]

Another image, this time of my fair-skinned granddaughter.  Left hand side is in the sRGB space with no conversion, right hand side converted to the iPad profile.  When viewed on my iPad 1, her face loses all color on the un-converted side, as well as her blouse being a paler pink, and there is some loss of detail in her hair.

One additional note - I created the profile with the iPad set to approx. 80% brightness.

Richard Southworth

[kimballistic]

Yeah, I'm seeing the same differences you describe.  The pic on the right has more detail in the shadows, although the increased saturation of her shirt seems to have erased its texture. 

[rasworth]

Good eye - I went back and checked the conversion from sRGB to iPad, this time using the View > Proof Setup menu in Photoshop to first do a softproof.  Sure enough, the entire blouse and a lot of the red in the background were out of gamut for the iPad, and therefore heavily clipped by the Relative intent conversion.

I went back to the raw file, juggled Vibrance and Saturation, trying to tone down the blouse but still keep some color in her face (she is quite light complexioned).  The result (attached) could use more work, but at least most of the colors are now in gamut.  Points up another good use for an iPad profile, i.e. make sure colors are going to come across without clipping.

Richard Southworth

[stefohl]

I know that Eizo is working on a solution where you can get their monitors to emulate an iPad or iPhone. If they will offer a profile for the different devices, so you could convert your image to that colour space, that I don't know.

[joofa]

Another image, this time of my fair-skinned granddaughter.  Left hand side is in the sRGB space with no conversion, right hand side converted to the iPad profile.  When viewed on my iPad 1, her face loses all color on the un-converted side, as well as her blouse being a paler pink, and there is some loss of detail in her hair.

One additional note - I created the profile with the iPad set to approx. 80% brightness.

Richard Southworth

I think the right image looks better.

[Todd Shirley]

Hey Richard

Can you post the actual icc profile you created so we can test it ourselves?

-Todd Shirley