PhotoSlop Compared to 4 Different Editors

  • Thread starter Thread starter Too Funny
  • Start date Start date
T

Too Funny

[corrected URL]

I thought it would be fun to add yet one more graphic editor into the
testing results, and then combine them all into one easy to see chart so
people don't have to bother clicking on 5 different links. Then trying to
remember what you saw at each one (I know how slow some of you are).


"Granger Calibration Chart" Editor-Test Results

http://farm3.static.flickr.com/2790/4337946696_8ef5e104ff_o.jpg

Isn't this fun? :-)
 
[corrected URL]

I thought it would be fun to add yet one more graphic editor into the
testing results, and then combine them all into one easy to see chart so
people don't have to bother clicking on 5 different links. Then trying
to remember what you saw at each one (I know how slow some of you are).


"Granger Calibration Chart" Editor-Test Results

http://farm3.static.flickr.com/2790/4337946696_8ef5e104ff_o.jpg

Isn't this fun? :-)

You can add this one I just created with GIMP. GIMP doesn't have
"Luminance" in its layer mixers, but applying "Hard Light" to the second
gradient seems to produce the same effect.

http://www.arumes.com/temp/GrangerChart.jpg

I also don't understand why the Luminance Landscape author says that the
background color must be white. The background will be gone after applying
the rainbow gradient anyway.

Ofcourse, it would be nice to see a response by a PS-user, who can say if
the error really is created by PS instead of the original author. I've
seen more stupid errors from LL before.
 
[corrected URL]

I thought it would be fun to add yet one more graphic editor into the
testing results, and then combine them all into one easy to see chart so
people don't have to bother clicking on 5 different links. Then trying
to remember what you saw at each one (I know how slow some of you are).


"Granger Calibration Chart" Editor-Test Results

http://farm3.static.flickr.com/2790/4337946696_8ef5e104ff_o.jpg

Isn't this fun? :-)

You can add this one I just created with GIMP. GIMP doesn't have
"Luminance" in its layer mixers, but applying "Hard Light" to the second
gradient seems to produce the same effect.

http://www.arumes.com/temp/GrangerChart.jpg

Thanks.

Yes, I had to use that same "Hard Light" layer-blend option in one of the
editors, I think it was in PhotoImpact. And another even more inexpensive
and obscure editor worked the same (PhotoScape? or something like that),
but the chart looked a little too different from the others to include it,
it was using some "Legacy Hard Light" method or something that clipped all
the blacks and lights. But even then the colors in that program, like
yours, were a nice even spread without all those horrendous hills and
valleys of the PhotoSlop one.
I also don't understand why the Luminance Landscape author says that the
background color must be white. The background will be gone after applying
the rainbow gradient anyway.

Not sure why. It could depend on if they have their system set to create
layers with a default transparency? Perhaps an easier way to circumvent
other settings. In either case, that's not even 1/10th his problem. :-)
Ofcourse, it would be nice to see a response by a PS-user, who can say if
the error really is created by PS instead of the original author. I've
seen more stupid errors from LL before.

When I spotted the difference on that page I got curious. That's why I
created that PhotoSlop Granger Chart on the comparison list I posted using
my own copy of PhotoSlop to see if he made any errors. He did not. I even
tried changing the system color profiles in PhotoSlop. I changed the
"rainbow gradient" to the true colors that they should be (they are way off
in PhotoSlop). They should be, from left to right, in 100% saturations:

Red, Magenta, Blue, Cyan, Green, Yellow, Red

At spacing increments of

0%, 16%, 33% 50%, 66%, 83%, 100%

Or if using hue-rotation degrees, then:

0, 60, 120, 180, 240, 300, 360

PhotoSlop's default "rainbow" gradient was all over the map in colors and
spacing. Further adding to any PhotoSlop user's nightmare when using this
method without correcting the gradient first. When I corrected PhotoSlop's
gradient even that didn't help. It resulted in a similar mess as what you
see in the above chart.
 
[corrected URL]

I thought it would be fun to add yet one more graphic editor into the
testing results, and then combine them all into one easy to see chart so
people don't have to bother clicking on 5 different links. Then trying
to remember what you saw at each one (I know how slow some of you are).


"Granger Calibration Chart" Editor-Test Results

http://farm3.static.flickr.com/2790/4337946696_8ef5e104ff_o.jpg

Isn't this fun? :-)

You can add this one I just created with GIMP. GIMP doesn't have
"Luminance" in its layer mixers, but applying "Hard Light" to the second
gradient seems to produce the same effect.

http://www.arumes.com/temp/GrangerChart.jpg

I also don't understand why the Luminance Landscape author says that the
background color must be white. The background will be gone after applying
the rainbow gradient anyway.

Agreed. Has no effect.
Ofcourse, it would be nice to see a response by a PS-user, who can say if
the error really is created by PS instead of the original author. I've
seen more stupid errors from LL before.

I get the same effect as on the LL page.

I tried the LL instructions in CS3 with various color settings, paper
profile settings (in simulate paper mode), etc. and I always get the
same general pattern at different tone levels.

What "error"? As long as the end result is correct, who cares how the G
chart looks? It's real purpose is to find dead or blocked areas, not to
create a linear look.

How can you find dead or blocked areas that don't exist on the PhotoSlop
Granger Chart?

[now waiting for that cartoonist's light-bulb to get drawn in above your
head]
 
I get the same effect as on the LL page.

I tried the LL instructions in CS3 with various color settings, paper
profile settings (in simulate paper mode), etc. and I always get the
same general pattern at different tone levels.

What "error"? As long as the end result is correct,

But is it? It this chart the correct result of combining two simple
gradient layers? Does that mean that the error is in _all_ the other
editors?
who cares how the G chart looks? It's real purpose is to find dead or
blocked areas, not to create a linear look.

That is the purpose of the Granger chart, and ofcourse I don't care how
that chart looks at all. But if I was a PS user, I'd be very worried about
the way PS combines two simple layers. That _does_ affect the end result.
 
Ofcourse, it would be nice to see a response by a PS-user, who can say
if the error really is created by PS instead of the original author.
I've seen more stupid errors from LL before.

When I spotted the difference on that page I got curious. That's why I
created that PhotoSlop Granger Chart on the comparison list I posted
using my own copy of PhotoSlop to see if he made any errors. He did not.
I even tried changing the system color profiles in PhotoSlop. I changed
the "rainbow gradient" to the true colors that they should be (they are
way off in PhotoSlop). They should be, from left to right, in 100%
saturations:

[...]

Havve you tried other blending modes in PS? Perhaps this is just the way
their Luminance blend is supposed to work.
If I convert my own chart to grayscale based on luminosity, I don't get an
even black to white gradient (which I do get with a Lightness grayscale).

So apparently a luminosity blend is different from a hard light blend. But
if I convert the PS version to grayscale based on luminosity, I also don't
get a correct black to white gradient.
 
You can add this one I just created with GIMP. GIMP doesn't have
"Luminance" in its layer mixers, but applying "Hard Light" to the second
gradient seems to produce the same effect.

http://www.arumes.com/temp/GrangerChart.jpg

I added yours to the comparison chart. Thanks. Had to change the
perspective a little to fit but that doesn't change the overall appearance.

Here's the link to the more complete version.

http://farm5.static.flickr.com/4046/4338229864_1f3767b261_o.jpg

There's room on that for 2 more editors now. Anyone else want to add their
favorite editor that's not already represented on it? :-)
 
Ofcourse, it would be nice to see a response by a PS-user, who can say
if the error really is created by PS instead of the original author.
I've seen more stupid errors from LL before.

When I spotted the difference on that page I got curious. That's why I
created that PhotoSlop Granger Chart on the comparison list I posted
using my own copy of PhotoSlop to see if he made any errors. He did not.
I even tried changing the system color profiles in PhotoSlop. I changed
the "rainbow gradient" to the true colors that they should be (they are
way off in PhotoSlop). They should be, from left to right, in 100%
saturations:

[...]

Havve you tried other blending modes in PS? Perhaps this is just the way
their Luminance blend is supposed to work.
If I convert my own chart to grayscale based on luminosity, I don't get an
even black to white gradient (which I do get with a Lightness grayscale).

So apparently a luminosity blend is different from a hard light blend. But
if I convert the PS version to grayscale based on luminosity, I also don't
get a correct black to white gradient.

Good thinkin'! I just tried using the Hard Light blend using PhotoSlop, and
it does indeed appear to work, most of the evidence of those grossly huge
valleys and peaks disappears. But the results look like crap. All the
colors are dulled and shifted. So does this mean they don't even know what
"Luminance" means and all the other programs do? :-) Would this also
explain why most all photos posted to the net by PhotoSlop users are so
grossly oversaturated because they have to over-compensate for each photo's
saturation on their own screen?
 
Robert said:
Too said:
Ofcourse, it would be nice to see a response by a PS-user, who can say
if the error really is created by PS instead of the original author.
I've seen more stupid errors from LL before.

When I spotted the difference on that page I got curious. That's why I
created that PhotoSlop Granger Chart on the comparison list I posted
using my own copy of PhotoSlop to see if he made any errors. He did not.
I even tried changing the system color profiles in PhotoSlop. I changed
the "rainbow gradient" to the true colors that they should be (they are
way off in PhotoSlop). They should be, from left to right, in 100%
saturations:

[...]

Havve you tried other blending modes in PS? Perhaps this is just the way
their Luminance blend is supposed to work.


Screen mode gives a symmetrical chart across the colors, as well as some
other modes, hard light is similar. Immediately after choosing a mode,
hit the up/down arrow keys. Not sure why luminosity works differently in
other programs to the PS luminosity mode but as Alan said, it's still
useful for their purposes or they wouldn't have put it out there,
obviously.

Luminosity means apply the lightness/darkness to the colors, and it's
apparently counting the luminosity of the rainbow layer also, so the
dark blues move down & the light yellows move up. This is the same sort
of issue where you have to be careful what method you use for converting
to b&w.
 
Robert said:
Too said:
Ofcourse, it would be nice to see a response by a PS-user, who can say
if the error really is created by PS instead of the original author.
I've seen more stupid errors from LL before.

When I spotted the difference on that page I got curious. That's why I
created that PhotoSlop Granger Chart on the comparison list I posted
using my own copy of PhotoSlop to see if he made any errors. He did not.
I even tried changing the system color profiles in PhotoSlop. I changed
the "rainbow gradient" to the true colors that they should be (they are
way off in PhotoSlop). They should be, from left to right, in 100%
saturations:

[...]

Havve you tried other blending modes in PS? Perhaps this is just the way
their Luminance blend is supposed to work.


Screen mode gives a symmetrical chart across the colors, as well as some
other modes, hard light is similar. Immediately after choosing a mode,
hit the up/down arrow keys. Not sure why luminosity works differently in
other programs to the PS luminosity mode but as Alan said, it's still
useful for their purposes or they wouldn't have put it out there,
obviously.

Luminosity means apply the lightness/darkness to the colors, and it's
apparently counting the luminosity of the rainbow layer also, so the
dark blues move down & the light yellows move up. This is the same sort
of issue where you have to be careful what method you use for converting
to b&w.

I guess this means anytime someone is using a lighten or darken
(dodge/burn) brush in PhotoSlop or creating any changes in luminosity, like
trying to recover details from shadows or highlights, they can expect all
their colors to get shifted just as badly. Nice! :-)

A good thing my copy was given to me or I wouldn't spend one cent on
something like this. The *ONLY* reason I even keep it on a computer at all
is for the one or two worthwhile plugins that are PhotoSlop dependent. Soon
as I use it for that plugin's effect and am done, it's shut down again with
the hopes that I don't have to ever run it again too soon. There's far too
many better programs out there than having to put up with its nonsense.
 
Too said:
Paul Furmanwrote:

I guess this means anytime someone is using a lighten or darken
(dodge/burn) brush in PhotoSlop

No, the burn & dodge tools work fine on the rainbow layer. It is an odd
behavior though. I sometimes use luminosity mode on adjustment layers,
rather than going to lab mode so that contrast/levels/curves adjustments
don't effect saturation. Hmm, so yeah this does make a difference. If I
try the granger thing in lab mode (set lab before adding layers) that
makes a very different chart with much softer transitions like the gamut
chart they describe in the LL article. Do these other programs have lab
mode? CMYK mode produces yet another different effect.

I checked adjustment layers on a couple normal photos, one with oranges,
one with orages & blues, both have green backgrounds. Lab vs
luminosity-mode adj layers had very slight color shifts making the
yellows brighter, no effect on the blues. Normal mode made the
saturation increase and orange showed more effect. So anyways, yeah, it
does seem to be a bit of a problem but if you don't like the color
shift, add another adjustment layer or use another program if you like.
I've been using photoshop for 20 years so it's what I like working with
and it has a whole lot of tools for whatever needs to be done.

Photoshop is not the tool for scientific experiments & measurement.
Roger Clark found some weird rounding errors & bit depth issues when
using it to make extreme adjustments to shadow detail in raw
conversions. I forget what program he used instead, something designed
for astrophotography manipulations I think. Not the sort of program I'd
like to use for general photo editing.

These days I mostly use Lightroom for tweaking exposures, contrast etc,
which also is not perfect but it's great for workflow. If I'm actually
going to do a large print of a difficult subject, I'll take the raw file
into my old-ish version of photoshop & do fussy stuff like sharpen a
luminosity layer, etc. Otherwise I don't use photoshop unless I need to
do some serious cloning or editing like that. Lightroom is quick and
effective, letting me concentrate on what's important. If a contrast
adjustment increases saturation, I'll just tweak the saturation slider.
If the oranges pop, I'll tune that, etc.
 
Alan said:
I've been thinking exactly the same thing. But then the simple truth
comes up: people have been printing WYSIWYG (close enough) from PS (and
the others long enough that the odd looking output of PS with those two
functions does not seem to matter very much. Or much. Or at all.


It is strange, to be sure. But not so much as a color issue as a layer
tool issue. I don't use layers much with gradients however.

I looked at the linearity of the B/W and spectrum gradients, and they seem
to be straight. So in mixing they should band straight in the vertical.
The diagonals are puzzling.

I've spent hours attempting to understand the seemingly odd result produced
by PS. I've either explained it or I went totally insane in the process
(yeee haa).

There are a few different ways of mapping RGB to HSB:
1. HSI intensity I=R/3+G/3+B/3
2. HSV value V=MAX(R,G,B)
3. HSL lightness L=0.5*MAX(R,G,B)+0.5*MIN(R,G,B)
4. Luminance Y'601=0.3R+0.59G+0.11B
5. and many others.

My old version of PS shows Hº S% B%, so what is B?
RGB 0,255,0 -> B 100%
2,70,29 -> 27%
70, 0, 0 -> 27%

I conclude B=100*MAX(R,G,B)/255. Of course, these are not linear values,
they are the 8-bit (insert colour profile name) encoded values.

Now, given our horizontal colour gradient with S=100% and H going from 359º
down to 0º left to right, what will happen when we attempt to change the
luminance from 0 at the bottom to 100% at the top?

Using Y'601 equation, we will get maximum percieved intesity of the colours
as follows:
Red around one third of the way up.
Green around two thirds of the way up.
Blue around a tenth of the way up.

That is what I see in PS and not in any of the others, therefore the others
are using a different equation for brightness (whatever each editor calls
it).

Any thoughts?

Pete
 
...
Now, given our horizontal colour gradient with S=100% and H going from 359º
down to 0º left to right, what will happen when we attempt to change the
luminance from 0 at the bottom to 100% at the top?
...
Red around one third of the way up.
Green around two thirds of the way up.
Blue around a tenth of the way up.
...

Over my head
....but that got me thinking about lab color space and there's some
interesting similar issues involved there:
http://en.wikipedia.org/wiki/Lab_color_space

In any case, I reproduced the chart like the others using two blending
modes, patched together:
http://www.flickr.com/photos/edgehill/4339700662/
see the other similarly named thread for that discussion.

Each of the blending modes does a different calculation like your
formulas. You can see a few different approaches in adjacent pics at
that link, including lab mode.
 
Paul said:
Over my head
...but that got me thinking about lab color space and there's some
interesting similar issues involved there:
http://en.wikipedia.org/wiki/Lab_color_space

In any case, I reproduced the chart like the others using two blending
modes, patched together:
http://www.flickr.com/photos/edgehill/4339700662/
see the other similarly named thread for that discussion.

Each of the blending modes does a different calculation like your
formulas. You can see a few different approaches in adjacent pics at that
link, including lab mode.

Thanks Paul.

I've been looking carefully at the Granger charts shown in this thread. For
the printer I use the PS chart shows the effect of limited gamut more
obviously than the others - meaning it takes less time to see the effect.

Certainly, the other editors are not performing a luminance blend, they must
be doing brightness, lightness, or intensity instead.

Anyone know where the original Granger chart can be found? I've nearly worn
out Google trying to find it (or just Granger's description of it).

Pete
 
Pete said:
I've spent hours attempting to understand the seemingly odd result
produced by PS. I've either explained it or I went totally insane in the
process (yeee haa).

Must be insanity - I'm replying to myself. It's an update :-)
There are a few different ways of mapping RGB to HSB:
1. HSI intensity I=R/3+G/3+B/3
2. HSV value V=MAX(R,G,B)
3. HSL lightness L=0.5*MAX(R,G,B)+0.5*MIN(R,G,B)
4. Luminance Y'601=0.3R+0.59G+0.11B

It's called luma.
5. and many others.

Using Y'601 equation, we will get maximum percieved intesity of the
colours as follows:
Red around one third of the way up.
Green around two thirds of the way up.
Blue around a tenth of the way up.

I've measured the result using a 1001 high image because it's much easier to
read off the values.

PS 5.0 LE gives the values:
luma = 0.3R + 0.59G + 0.11B

which means it's actually using ITU-R Rec. BT.601 NTSC
luma Y'601 = 0.299R+0.587G+0.114B

for sRGB it should be ITU-R Rec. BT.709
luma Y'709 = 0.2126R+0.7152G+0.0722B

There are many reasons for not using ITU 709, so 601 is probably fair
enough.

=================
The other editors
=================
The version of GIMP I saw does not even have a Luminance blend mode.

PSP is confusing because later versions have two options: Luminance and
Luminance Legacy - one uses lightness the other uses luminance. Descriptions
say neither affect colour. Then it isn't using luma - luma will affect
colour.

I didn't bother looking at any of the others because its's obvious why they
all differ from PS. It seems they all used a reference such as this
http://en.wikipedia.org/wiki/Blend_modes#Luminosity
which is wrong. It says "Luminosity takes its brightness... Top: B" WTF?
Then call it brigtness blending because it isn't luma.

PS is using the Luma/Chroma/Hue Model of polar-coordinate representation,
which is precisely defined for RGB in computer graphics and digital TV.

The other editors appear to be using the HSL "Double Hexcone" Model
therefore the blending mode is lightness, hence no colour shift - they
cannot produce the Granger Chart because they do not have a luma blending
mode.

This thread has shown us Confirmation Bias working at its best.
http://en.wikipedia.org/wiki/Confirmation_Bias

Pete
 
I am going to jump into this discussion with an idea, not based upon any
direct knowledge I have as to why Photoshop is designed as it is, but
with a mind toward human vision.

Human vision is not linear at all. We see by two structures, rods, which
perceive luminosity without color, and cones, three forms of which each
see a different light frequency range, the "red" cones actually peak at
yellow, green cones peak at green and blue cones peak in the violet part
of the spectrum. These structures are not evenly distributed in terms
of numbers or locations. Rods are more concentrated at the edges of our
retina and are used in our peripheral vision, which has poor color
vision. Cones are more concentrated in the middle of the retina.

While the percentages of the three cone cells vary considerably between
adults, thus creating the various levels and forms of color blindness,
overall adults have very poor blue-violet color perception. We are born
with less blue cones than either the red or green. Early in life,some
are destroyed by UV light entering the eye. The UV also damages the
transparent and neutral lens of the eye, slowly yellowing it until it is
almost orange juice color as we pass middle age. While this yellowing
actually filters the UV light reaching the retina, and therefore
protects the blue cones left, it also filters out a huge amount of the
blue light that reaches the retina, so a mixture of low blue cones to
begin with, damage by UV early in life, and then a fairly heavy yellow
orange filtration, means our blue perception stinks.

While we are very attentive to reds, due to the evolutionary
significance of red to yellow, it is the second smallest number of
cones. So, while we are more aware of reds, we actually have less red
cones to see it with. Our largest number of cones are greens, in most
people by a substantial amount. This allows us to see green
differentials quite well, and is probably significant because vegetation
coloration can be important to survival.

Now, I'm not sure what Photoshop is up to with their Granger Calibration
Chart differences (yet at least - I've asked a friend who is a color
engineer if he had any comments) but as someone who has a background
with the human side of color perception, it may be that Adobe has made
some accommodation for the species who use their programs, with our
non-linear color perception.

Art




If you are interested in issues surrounding e-waste,
I invite you to enter the discussion at my blog:

http://e-trashtalk.spaces.live.com/
 
Robert said:
If I convert my own chart to grayscale based on luminosity, I don't get an
even black to white gradient (which I do get with a Lightness grayscale).

Oh, I see what you mean. The odd looking asymmetrical PS version, if
converted to grayscale, goes to a perfect gradient black on bottom,
white on top. The others show dark & light vertical shapes like the
color version. BTW I did a version in lab mode and had to convert to RGB
for web... that rgb version looks very similar but doesn't go back to a
smooth gradient but if I flatten the psd and convert that to grayscale
while still in lab, it works. The lab version takes out the weird
angular components, apparently lab is designed to transition more
smoothly & match human vision better.

None of these gives the appearance of a 'true' parallel set of vertical
rainbow bands fading from black to white which one might expect from the
exercise. Optics can be similarly baffling where you correct for
spherical aberration, certain kinds of chromatic aberrations get worse
or the bokeh goes ugly, etc.
 
Back
Top