Brilliancy vs Brightness

[flyingpig]

these 2 images have exactly same objective average brightness. Do they really have similar Brilliancy for you?


To calculate Brilliancy metric we use Brightness, Contrast, Pattern (static and dynamical pattern). just average and contrast are not enough to receive good correlation with Human score .

in same time first diamond( MSS13, score 1.13) has more darker black areas than second diamond.
Please compare main facets under table.
when white is brighter and black is darker then the contrast is higher. ( with same average brightness)

In this case, since two stones have the same average brightness, whichever has higher contrast will be more brilliant because white will be whiter and brighter.

Having that said, a diamond can appear brilliant in the absence of contrast within it.

 

[drk14]

@Serg, my recommendations to you would be:

(1) Present randomly drawn pairwise comparisons to the test subject (not the full array of 30 stones), so that the subject is only asked to determine the relative rank of two images at a time. You can then process these data using standard analysis methods for paired comparisons (BTL, etc.) to determine the underlying scale.

(2) Provide clear instructions to the subject (e.g., "which diamond do you prefer?", "which diamond appears brighter to you?", etc.). Everybody has a different definition of "brilliance", so in my opinion, you well get a lot of noise in the data if you ask "which diamond is more brilliant?".

 

[Serg]

In this case, since two stones have the same average brightness, whichever has higher contrast will be more brilliant because white will be whiter and brighter.

Having that said, a diamond can appear brilliant in the absence of contrast within it.

I never saw diamond without any contrast in real light environments . You need perfect white integral sphere with pinhole to receive a diamond without contrast . In real light environments diamond always has at least random 1st reflection from facets. Such big reflections create a brilliancy.
Also diamond usually has strong contrast with background

 

[Serg]

@Serg, my recommendations to you would be:

(1) Present randomly drawn pairwise comparisons to the test subject (not the full array of 30 stones), so that the subject is only asked to determine the relative rank of two images at a time. You can then process these data using standard analysis methods for paired comparisons (BTL, etc.) to determine the underlying scale.

(2) Provide clear instructions to the subject (e.g., "which diamond do you prefer?", "which diamond appears brighter to you?", etc.). Everybody has a different definition of "brilliance", so in my opinion, you well get a lot of noise in the data if you ask "which diamond is more brilliant?".

early( 3-5 years ago ) we used such approach. It does not work well if many samples have very narrow difference in metric . It works fine if you have to sort 10-20 sample with well marked difference in performance . it could be helpful tool for consumers to sort them wishing list .

 

[Serg]

In your 2013 article, you include 10 different definitions of brilliance (Box A), and you also make the claim "We know that neither a static image (a photograph) of a diamond on a monitor, nor its printed image has brilliance" (page 89). So I guess I will need a little more explanation about what criterion you want us to use to sort the diamond images.

Current monitors have higher dynamical range and resolution than 5-7 years ago. it is important to see Brilliancy( Illusion) on monitor . Modern TV gives even more opportunity to see brilliancy on static image . there is not any visible Brilliancy on printing mono images. But you may see Brilliancy on printed stereo images .
Roughly 90% Brilliancy come from tilting and stereo.

 

[drk14]

early( 3-5 years ago ) we used such approach. It does not work well if many samples have very narrow difference in metric .

This suggests that you can solve the problem by recruiting a larger number of test subjects.

But if your goal is just to calibrate and/or validate your proposed metric, I think you should be able to do this using a subset of images that are sufficiently well separated.

 

[arkieb1]

O.K this is a screen shot taken with an Iphone of what I see on my computer, I took both images and put one on top of the other - compare the emerald in both, so in your first pic where you asked us if stones with similar brightness have a different order of brilliancy I would argue yes at home on my screen they do, when you cut and pasted another static pic of the 4 stones I can clearly see the difference in brilliancy and in the videos but when I compare that back to your original chart here is a screen pic of what I actually see. Hopefully using the emerald as an example you can see the problem we are having at home.

 

[Serg]

O.K this is a screen shot taken with an Iphone of what I see on my computer, I took both images and put one on top of the other - compare the emerald in both, so in your first pic where you asked us if stones with similar brightness have a different order of brilliancy I would argue yes at home on my screen they do, when you cut and pasted another static pic of the 4 stones I can clearly see the difference in brilliancy and in the videos but when I compare that back to your original chart here is a screen pic of what I actually see. Hopefully using the emerald as an example you can see the problem we are having at home.

In 1st post I published 2 tables:
1) 1st table with roughly equalized average brightness ( averaged by pixels brightness)


The Emerald has brightness 166(160) and Brilliancy score 1.01

2) 2d Table with original brightness


The Emerald has brightness 129 and Brilliancy score 0.68


Movies from link
https://cutwise.com/~gvC3

also have original brightness.
Of course a difference in brilliancy is much bigger without brightness equalization.
And of course a video comparison shows clearly a difference between emerald cuts and other cuts

 

[valeria101]

these 2 images have exactly same objective average brightness. Do they really have similar Brilliancy for you?
ATTACH #507893
ATTACH #507894
To calculate Brilliancy metric we use Brightness, Contrast, Pattern (static and dynamical pattern). just average and contrast are not enough to receive good correlation with Human score .

Of course not.

The right-hand stone looks like sapphire - one with foggy Crystal (@Sagebrush term)

I would trade alot of Brightness for Contrast, especially if the Pattern is a sufficiently fine partition of the viewer-facing area.

This 'sufficiently fine' is a characteristic of various cut models for me - I know I have a dyed in the wool preference for unusually broad ones, but tend to appreciate achievement within each type - as acquired tastes ...

 

[Texas Leaguer]

In this case, since two stones have the same average brightness, whichever has higher contrast will be more brilliant because white will be whiter and brighter.

I believe that the perception of brilliance also depends on the overall proportion of contrast in a diamond as well as its distribution.

 

[Texas Leaguer]

Not sure it is possible to decouple brilliancy from scintillation, which would factor in not only dynamic contrast, but size of virtual facets, frequency, and pattern.

 

[gm89uk]

Hi Serg,

This work reminds me of pattern reversal visual evoked potentials (VEP), which involve (rough definition), measuring cortical signal at the back of the head (near the visual cortex) based on an alternating checkerboard pattern. An alternating checkerboard pattern is synonymous with scintillation.

It is well established that reduced contrast, elicits a different (or reduced) response. You may want to involve a neuro electrophysiologist in giving an opinion on your metric (I am not one).

I have some concerns with your metric. Consider the diagram below, which have all equal luminance

Your metric is simply stating that the checkerboard on the left is more brilliant than the one on the right.
If you rearranged the brightness adjusted diamonds diagram by which one has the whitest pixel, instead of by your brilliance metric, you'd probably end up with the same result. If you correct average brightness, then the whiter diamonds, will have the darker patches, and be similar to the left checkerboard pattern (as flyingpig stated). The brightness adjusted diamonds that score poorly will have simply have less contrast. By adjusting luminance you are by definition only scoring contrast.

This is known to elicit a less impressive response in pattern reversal VEPs. You may find this an interesting read (https://link.springer.com/article/10.1023/A:1002750719094)

I agree however with your definition of brilliance:
"Diamond brilliance is an illusion caused by the fact that the perceived brightness of the object significantly exceeds its actual brightness. "
Should you then not be comparing something like the below:

They are adjusted to have the same peak brightness (the white squares), but reduced contrast in the left. The checkerboard on the right has higher luminance, but the brightest area of both diamonds are identical, unlike in your brightness adjusted ranking above (but this would probably end up being a measure of contrast as well).

I'm not entirely sure how you could assess brilliance (other than simple contrast measurement) on a static measurement!

2) Your ranking doesn't account for the stereoscopic effect of one eye seeing dark, another eye seeing bright in the same simultaneous spot of diamond, and the brain combining both to an illusion of brilliance, that you and Garry have numerously discussed, but not sure how this would be possible in a metric.

All the best

 

[Serg]

Hi Serg,

This work reminds me of pattern reversal visual evoked potentials (VEP), which involve (rough definition), measuring cortical signal at the back of the head (near the visual cortex) based on an alternating checkerboard pattern. An alternating checkerboard pattern is synonymous with scintillation.

It is well established that reduced contrast, elicits a different (or reduced) response. You may want to involve a neuro electrophysiologist in giving an opinion on your metric (I am not one).

I have some concerns with your metric. Consider the diagram below, which have all equal luminance

Your metric is simply stating that the checkerboard on the left is more brilliant than the one on the right.
If you rearranged the brightness adjusted diamonds diagram by which one has the whitest pixel, instead of by your brilliance metric, you'd probably end up with the same result. If you correct average brightness, then the whiter diamonds, will have the darker patches, and be similar to the left checkerboard pattern (as flyingpig stated). The brightness adjusted diamonds that score poorly will have simply have less contrast. By adjusting luminance you are by definition only scoring contrast.


All the best

Hi Gm,

We do not change a brightness to calculate a brilliance .
We changed the brightness to create samples for tests. It is important to verify Brilliance metrics on samples with same average luminance.
in Subjective Brightness metric we account local and global contrast. usually subjective brightness more depends from local contrast than from global contrast. Local contrast depends from pattern distribution and pattern size. Local contrast is not just the difference between min and max Luminance

 

[Serg]

Should you then not be comparing something like the below:

They are adjusted to have the same peak brightness (the white squares), but reduced contrast in the left. The checkerboard on the right has higher luminance, but the brightest area of both diamonds are identical, unlike in your brightness adjusted ranking above (but this would probably end up being a measure of contrast as well).



All the best

the result depends from checkerboard square size.
it will different for below samples.
btw which is most bright for you?

 

[Serg]

the result depends from checkerboard square size.
it will different for below samples.

 

[OoohShiny]

I love the science in this thread