Samsung Makernote 0x0020 SmartAlbumColorInfo

[Klaus_Homeister]

Hi Phil

This is a Gimmick from EX1 and some other compacts wich have an internal ImageBrowser called 'SmartAlbum'.
With that Browser you can do various sorting and one is sort by color.
And for only this Function 0x0020 stores the Info.

First Short is for a category.
Valid 0-7: 
0:Neutral
7:Other Color
even numbers 2/4/6 cool-colors  (to blue)
uneven numbers 1/3/5  warm colors (to red)

Second short is obviously used to fine-tune the sorting within the color-group.

Most Cams don't have this color-browsing and then both values are 0 (neutral position Black).

Greetings
Klaus




Tag 0x0020 SHORT      × 2       #000600A2 SmartAlbumColorInfo  Coolest (fine   0 | bright +162)

[Phil Harvey]

Thanks!  I'll post back after I get a chance to add these if I have any questions.

- Phil

[Phil Harvey]

Hi Klaus,

I was trying to find names for the colours.  I find some correlations, but nothing definite.  Looking at my Samsung samples, I see this:

0, 0: random (no smart color info?)
0, 32000-50000: most of these pictures have red, orange or skin tones in them
1, 29000: yellow
2, 29000-34000: green
3, 19000-48000: blue
4, 20000-22000: magneta
5, 37-115: black&white?
6: 160-300: gray?
7: 80-150: multicolored?

It would be useful if you could suggest some names for these color values.

Also, it would be interesting to have a better idea of what the second number means.  It is interesting that these numbers are much smaller for categories 5-7 for some reason.

- Phil

[Klaus_Homeister]

Hi Phil

I have lost the document, but there is an older patent from Samsung that explained this sorting by color-temperature (and not by RGB-Color).
'Green' 'Magenta' 'Yellow' don't fit to Color-temperatures.

In this Patent I remember:
0 was 'Neutral'  (and that was 'not available'  too)
7 was 'etcetera' (and that was 'out of range' too)

and then the valid range of the metering-system was split into below and above the calibrated default-daylight-temperature of a cam.
That's around 5000° - 5750° Kelvin for nearly all Cams and manufacturer but I don't know the exact Temperature for Samsung-Cams.

So my notation is:
Split Value into two shorts
The upper one is the smaller value range 0-7 and the 'main-index':
6 : ("Coolest")
4 : ("Cooler")
2 : ("Cool")
0 : ("Neutral/defaultDaylight")
1 : ("Warm")
3 : ("Warmer")
5 : ("Warmest")
7 : ("etcetera")

The lower short may be two separate bytes where the first would fit to brightness in my EX1-Shots. It is '0' with 'forgot the cap on the lens' and get >180 in bright overexposed shots.
The second Byte should be the fine-sort within the given Temperature-range ... but I have not found a system. 
Always wehen you're sure to 'have the system' ... then you find a series of shot's that do the opposite.
So ignore the second Short as 'unknown'. 

'coolest'  -> EX1 outdoor-shots in snowy winter
'neutral'  -> EX1 outdoor-shots at high-noon in late spring with clear sky or Greycard-Shot's
'Warmest' -> EX1 indoor-shots of a family-party in poor tungsten-lightning

'etcetera' is for example a EX1 shot where half of the image is indoors at tungsten-light,
and the other half of the image is the big window with overcast sky.
That is a situation where a correct metering isn't possible and 'etcetera' means 'Metering-Error'. 
Another Image is a pretty sundown with red purple magenta colored sky over nearly black landscape.
The twilightzone of light - short before sun is completely gone.
That is a situation where the scene is out-of-range for are usual Metering-Sytem. 
And yes ... a 'multicolored' Gretag-ColorCard.

We don't have enough information to make more precise Infos.

Klaus

[Phil Harvey]

I appreciate this detailed description, although from what I have seen it seems as if the patent may be unrelated to the feature implemented in the camera.  For example, the DV100 user manual says "View files by the dominant color of the image", which corresponds to what I observe with the sample images I have from many Samsung models.  (eg. all images with a 3 for this tag have a lot of blue in them -- typically daytime pictures with a lot of sky, and definitely not very warm.)  My samples don't seem to correlate well with the color temperatures you have listed.

- Phil

[Phil Harvey]

For example, here is a set of EX1 images organized by smart album color:

Smart album color 0 (red/orange?):



Smart album color 1 (yellow?):



Smart album color 2 (green?):



Smart album color 3 (blue?):



Smart album color 5 (black?):



Smart album color 6 (gray?):



Smart album color 7 (various?):



- Phil

[Klaus_Homeister]

Ok! You are right!

They're from Photography-Blog and I have downloaded some for testing.
The first of your Green pictures is shown om a green Patch of the Browserbar and that's the first time I have seen that.

I had an better idea - took my EX1 and in Macro-Mode and made a full-image-shot of every single color-patch one by one.

And then I understood. All my Photos are RAW only and underexposed fix -2. This way RAW's get perfect but embedded JPGs are seriously dark.
The measurement is done by those embedded Jpg's and always end up in reduced color-categories.

Now I have done a well exposed Series and my colorbar gets multicolored! You are right!
Directory screenshot with thumbs is attached.

And that is the Output of exiftool for those images:
I have added the PatchNo. and named colors from
https://en.wikipedia.org/wiki/ColorChecker

6 210   SAM_9768.JPG  Patch #19 White
6 210   SAM_9768.SRW 
6 190   SAM_9769.JPG  Patch #20 Neutral 8
6 190   SAM_9769.SRW 
6 158   SAM_9770.JPG  Patch #21 Neutral 6.5
6 159   SAM_9770.SRW 
5 112   SAM_9771.JPG  Patch #22 Neutral 5
5 112   SAM_9771.SRW 
5 69    SAM_9773.JPG  Patch #23 Neutral 3.5
5 69    SAM_9773.SRW 
5 29    SAM_9774.JPG  Patch #24 Black
5 29    SAM_9774.SRW 
3 32845 SAM_9775.JPG  Patch #13 Blue
3 32845 SAM_9775.SRW 
2 32767 SAM_9776.JPG  Patch #14 Green
2 32767 SAM_9776.SRW 
0 34449 SAM_9777.JPG  Patch #15 Red
0 34449 SAM_9777.SRW 
1 32767 SAM_9778.JPG  Patch #16 Yellow
1 32767 SAM_9778.SRW 
4 32768 SAM_9779.JPG  Patch #17 Magenta
4 32768 SAM_9779.SRW 
3 32767 SAM_9780.JPG  Patch #18 Cyan
3 32767 SAM_9780.SRW 
1 32767 SAM_9781.JPG  Patch #7 Orange
1 32767 SAM_9781.SRW 
2 16551 SAM_9782.JPG  Patch #8 Purplish Blue
2 16551 SAM_9782.SRW 
4 32767 SAM_9783.JPG  Patch #9 Moderate Red
4 32767 SAM_9783.SRW 
0 32767 SAM_9784.JPG  Patch #10 Purple
0 32767 SAM_9784.SRW 
3 37200 SAM_9785.JPG  Patch #11 Yellow Green
3 37200 SAM_9785.SRW 
1 22204 SAM_9786.JPG  Patch #12 Orange Yellow
1 22204 SAM_9786.SRW 
0 40366 SAM_9787.JPG  Patch #1 Dark Skin
0 40366 SAM_9787.SRW 
0 33173 SAM_9788.JPG  Patch #2 Light Skin
0 33173 SAM_9788.SRW 
3 34795 SAM_9789.JPG  Patch #3 Blue Sky
3 34795 SAM_9789.SRW 
2 32767 SAM_9790.JPG  Patch #4 Foliage
2 32767 SAM_9790.SRW 
4 32767 SAM_9791.JPG  Patch #5 Blue Flower
4 32767 SAM_9791.SRW 
3 31530 SAM_9792.JPG  Patch #6 Bluish Green
3 31530 SAM_9792.SRW 

I would say now:

0 Red / Purple / Dark Skin / Light Skin
1 Yellow /  Orange / Yellow Green / Orange Yellow
2 Green / Purplish Blue /  / Foliage
3 Blue / Cyan / Blue Sky / Bluish Green
4 Magenta / Blue Flower
5 Black / dark Grey
6 White / light grey

On the Image-Browser for pure colors the bar is colored ..
0 Red
1 Yellow
2 Green
3 Blue
4 Purple
5 Black
6 Grey
7 The complete Colorchecker isn't attached but gets an 7 and shows an Text 'Etc.' on a white patch

For example
The File 3 32767 SAM_9780.JPG  Patch #18 Cyan
is exactly on the border of the solid Green on left side and solid Blue on right side

maybe the 32767 is a signed short -1 and indicates a 'left border' of the BluePatch ?

But all the neutrals SAM_9768 > 9774 Patch #19 > #24 ...
have values 29 from #Black ... 69/112/158/190 up to #white 210

That looks like the brightness of the color Black and color Grey ... that would be the saturation of a color.
We have here the lowest and highest saturation in Black and white ... values 29 and 210
-> That's a BYTE
Last Byte maybe is saturation.

Let's look to Cyan Patch 32737 That's 0x7FFF -> splittet two bytes 127 255
Last one maximum saturation of 255
First one hue-shift to the most left side of the current category ?
That would fit all the neutrals with first byte 0 -> center of the color and no hue-shift

NO:
2 32767 SAM_9776.JPG  Patch #14 Green
is 32767 like Cyan but is exactly in the middle of Green-Bar and both borders far away

Hmmh
0 34449 SAM_9777.JPG  Patch #15 Red
0 40366 SAM_9787.JPG  Patch #1 Dark Skin

They're side by side in the middle of the red-colored part of the bar.

Red            34449 -> 0x8691   ->  134 / 145
Dark Brown 40366 -> 0x9DAE  ->  137 / 174

DarkBrown is the darkest version of red but has the second byte higher than the pure Red!
So the second byte can _not_ be the brightness/saturation/intensity or what ever.
Also the hue-shift can't be correct ...
dark brown and pur intense red should be seriously different in hue ... the first byteval 134<>137 ... next to next

I give up!

We have the first short fitting your RGBCMY/black/grey/others
That looks good, makes sense and has logics

The second Short has nothing of that! It's for sorting within the color-category and only the cam knows the way it's done.

And exiftool should check for a LONG 0
The lowest value I can get with EX1 is '16' for the second short and then category is 5/black.
with cap on the lens and underexpose to lowest possible exposure -2.0EV.
So if the Cam supports this Tag, it's _never_ completely null. If not it is _always_ null.

Klaus

PH Edit: Removed blank lines

[Phil Harvey]

Very interesting, thanks for looking into this!  I had thought about saturation for the 2nd number too, but discounted it because it didn't correlate with my samples.  For now I'm happy just printing it as a number.

- Phil