Metrics of beauty, BScope pics vs averaging pixels

[michaelgem]

This is a resurfacing of a dead thread.

Consider that metrics for diamond beauty that count pixels to obtain average or weighted measures of brilliance, fire etc. loose the detailed information available from snapshots such as those that Barry asks you to look at.

Tim and Barry,

Re Tim''s statement: "Essentially, the BScope moves a light source along a predetermined arc across the face of the diamond, and takes digital images of the diamond at five predetermined positions along that arc."

As someone very familiar with the workings of the BScope, I am cognizant of its appeal and its shortcomings.

As an important point of correctness, the ring light source does not move along an arc across the face of the diamond. It simply moves along the axis perpendicular to the diamond''s table, between the camera lens and the diamond. The lens shoots the diamond''s pics along this same axis through the center of the ring light.

This is important to understand, because this system evaluates the diamond only in the face-up position with a radially symmetrical light source. Many argue, rightly or wrongly, that a measure of brilliance, fire or scintillation should include viewing angles other than just this one salient viewing angle. Others say realistic light sources are not symmetrical.

I think this comment by Barry is particularly insightful:

Barry''s statement: “Insofar as the BrillianceScope is concerned, I would reiterate that consumers should primarily focus on the 5 light images generated by the machine.
It is here that one can assess the amount, intensity, and distribution of light emanating from the diamond.”

Barry''s exhortation to focus on the 5 snapshots accords with my research reported in "Diamond Brilliance" at www.acagemlab.com. Look at the ‘view more highlights’ section: WLR Averaging Verses ''Snap Shot" Analysis of Brilliance. http://www.acagemlab.com/target/index.htm

Here I point out that averaging to measure the diamond''s brilliance
"has lost the detailed knowledge of the relative brilliance occurring at any particular angle of observation such as the most important face-up position. Additionally, WLR has averaged out and lost the detailed information of the relative brilliance emitted at each point across the diamond''s surface. ''Snapshots'' retain this detail and can individually be used to assess brilliance at each point on the diamond and each viewing angle."

Check it out and let me know if you agree, or did I just confuse the issue for you?

Michael Cowing
www.acagemlab.com
http://www.acagemlab.com/target/index.htm

 

[Serg]

Hi, Michael
Yes, the radially symmetrical light source and one diamond viewing angle are big problems of BS.
Other can be compensated by software.

 

[Rank Amateur]

How big a role does the placement of the stone in the machine play in the apparent symmetrical-ness of the WLR? Does one move the stone around so as to return the highest score? Maybe the highest averaged score comes when the diamond is not perfectly placed in the center.

I've often wondered what the significance is of non-symmetrical returns displayed in the various light positions.

R/A

 

[Iiro]

Hi Michael,
it may take a while before I have completed to read your article.
Its huge !
S

 

[dbretton]

Hello MichaelGem,

Many argue, rightly or wrongly, that a measure of brilliance, fire or scintillation should include viewing angles other than just this one salient viewing angle.

In the URL you provided, you mention a study by Hemphill in which they utilized a cos^2 weighting function to weight the WLR readings. This rather plainly states that off-axis WLR readings are significant to a great degree.

Incidentally, how was such a weighting function derived? Common sense seems to indicate that this is correct, as many observers of a diamond do not view it from more than ~60 deg off normal. However, I was wondering what the scientific basis for this weighting function was.


You state in your post, and on the referential URL, that the BrillianceScope (BS) moves along the normal to the diamond table.
Does the BS identify the normal angle, or is this a task which the user must perform? If the prior is true, how is this accomplished?
Is the diamond placed on a mounting which is rotated until the critical angle to the table is found by transmitting a single wavelength pulse off the table?

"has lost the detailed knowledge of the relative brilliance occurring at any particular angle of observation such as the most important face-up position. Additionally, WLR has averaged out and lost the detailed information of the relative brilliance emitted at each point across the diamond's surface. 'Snapshots' retain this detail and can individually be used to assess brilliance at each point on the diamond and each viewing angle."

If the snapshots are significant because detailed information is lost in the averaging, then that is indicative of too small a sample set.

After reading this, I am further lost in one aspect: I still do not know why so few samples are taken when producing a BS reading. Everything posted seems to confirm that more images are needed. ?

 

[michaelgem]

Dennis: In the URL you provided, you mention a study by Hemphill in which they utilized a cos^2 weighting function to weight the WLR readings. This rather plainly states that off-axis WLR readings are significant to a great degree.

Michael: Actually, the cos square function drops off rapidly indicating GIA's desire to mostly weight the face-up normal viewing angle and relatively small tilts from the normal.

Dennis: Incidentally, how was such a weighting function derived? Common sense seems to indicate that this is correct, as many observers of a diamond do not view it from more than ~60 deg off normal. However, I was wondering what the scientific basis for this weighting function was.

Michael: The use of the cosine squared function makes little sense to me. This weighting function was an ad hoc attempt to heavily weight the face-up view and small tilts and almost eliminate the influence of large tilts such as the 60 deg one you mention. It has no more scientific basis than that.

Dennis: You state in your post, and on the referential URL, that the BrillianceScope (BS) moves along the normal to the diamond table.
Does the BS identify the normal angle, or is this a task which the user must perform? If the prior is true, how is this accomplished?

Michael: The camera lens is below a glass plate and faces up normal or perpendicular to it. Placing the diamond table down on the glass establishes the perpendicularity.


"has lost the detailed knowledge of the relative brilliance occurring at any particular angle of observation such as the most important face-up position. Additionally, WLR has averaged out and lost the detailed information of the relative brilliance emitted at each point across the diamond's surface. 'Snapshots' retain this detail and can individually be used to assess brilliance at each point on the diamond and each viewing angle."

Dennis: If the snapshots are significant because detailed information is lost in the averaging, then that is indicative of too small a sample set.
I still do not know why so few samples are taken when producing a BS reading. Everything posted seems to confirm that more images are needed.?

This logic indicates that you may have read my words but I failed to communicate.

1. The BS scope evaluates only one viewing angle, the face-up, normal viewing angle.

2. Randy Wagner has stated that 5 snapshots from this one viewing angle under ring light illumination from 5 distances are sufficient to grade the brilliance, fire and scintillation of a diamond.

3. Most feel that other viewing angles should be averaged in.

4. I feel that averaging looses information.
The single value of WLR or the BS brilliance measure obtained by an averaging of viewing angles or lighting angles has lost the detailed knowledge of the relative brilliance occurring at any particular angle of observation at any particular spot on the diamond..

5.'Snapshots' retain this detail and can individually be used to assess brilliance at each point on the diamond and each viewing angle.

6. By analyzing a 'snapshot' of a diamond, more can be learned about aspects of brilliance that are lost in a single measure that averages many viewing positions.

7. Because diamonds are evaluated for beauty in the face-up viewing position, the brilliance at this 'normal' viewing angle is of paramount importance. The principal concern in deciding to use snapshots, especially this 'normal' snapshot, for brilliance, instead of averaging viewing angles, is whether a diamond will retain its brilliance when it is tilted slightly from the perpendicular.

8.Our experience has shown that diamonds with near 'Ideal' proportion parameters maintain superior brilliance when viewed at angles off the perpendicular. Brilliance is essentially undiminished through 15° of tilt, only slightly at 25° and slightly more at 45°.

9. From all of this one may rightly conclude that superior brilliance from the face-up normal viewing angle is a necessary condition for an Ideal cut diamond. Because it has been shown by many over decades of analysis that Ideal cut diamonds retain their superior brilliance when tilted, one may further conclude that superior brilliance from the normal viewing angle is both a necessary and sufficient condition for an Ideal cut diamond.

10. Averaging pixels from normal view snapshots, as the BS does to obtain its measures, looses important aspects of brilliance such as our much discussed aspect of contrast brilliance and the even or uneven distribution of that brilliance across the viewed surface of the diamond.

If this has not put you to sleep, perhaps you will reread my brilliance article in light of what I have said here. Most likely only one in one hundred will even get this far. As my wife has said to me: “You must be mistaking me for someone cares about all this stuff.”

Michael Cowing
www.acagemlab.com
http://www.acagemlab.com/target/index.htm

 

[dbretton]

Michael: Actually, the cos square function drops off rapidly indicating GIA's desire to mostly weight the face-up normal viewing angle and relatively small tilts from the normal.


Actually, the cos square function doesn't drop off that rapidly. (cos (pi/6))^2 is still about 75% of the value at normal value. At about pi/3, the value is down to 25% of normal.


Dennis>>
Michael: Your common sense differs from mine. This weighting function was an ad hoc attempt to weight the face-up view and small tilts heavily and almost eliminate the influence of large tilts such as the 60 deg one you mention. It has no more scientific basis than that.

My common sense originates from doing a bit of professional work in the area of single variate and multivariate gaussian distribution functions over various problem spaces. My initial thoughts about how to approximate the weighting of the WLR were to use a gaussian distribution. Once I read your blurb about the cos^2 function, my reaction was that it was a semi-arbitrary fudge factor. I guess it was.

 

[michaelgem]

Michael: Actually, the cos square function drops off rapidly indicating GIA's desire to mostly weight the face-up normal viewing angle and relatively small tilts from the normal.

Dennis: Actually, the cos square function doesn't drop off that rapidly. (cos (pi/6))^2 is still about 75% of the value at normal value. At about pi/3, the value is down to 25% of normal.

Michael: Point taken.
Recognizing the importance of the face-up viewing position GIA stated they "wanted the contribution from rays that emerged straight up to be much greater."
As you rightly point out the cos squared function fails to accomplish this stated objective.

Dennis: My common sense originates from doing a bit of professional work in the area of single variate and multivariate gaussian distribution functions over various problem spaces. My initial thoughts about how to approximate the weighting of the WLR were to use a gaussian distribution. Once I read your blurb about the cos^2 function, my reaction was that it was a semi-arbitrary fudge factor. I guess it was.

Michael: Sorry about that. I should have shown more tact. I was indicating that use of the cosine squared function made little sense to me rather than seeming correct.

This function was indeed "a semi-arbitrary fudge factor."

I appreciate your rapid response and interest in this discussion, and apologize for my lack of tact.

What do you think about the numbered points?

What do you think of my "Diamond Brilliance" article's proposition that:

If the weight given to light rays near 90° were increased to the limit relative to other angles of observation, it would lead to a 'snapshot' of a diamond in the face-up position. Let us consider the suitability of a measure of brilliance using this and other 'snapshots'. Evaluating 'snapshots' of brilliance, especially the face-up, 'normal' viewing position has three important advantages over averaging.

Michael Cowing
www.acagemlab.com

 

[Serg]

The person usually does not use simple average function at the analysis of the preferences.
One parameter can reject a choice.

 

[michaelgem]

Sergey :The person usually does not use simple average function at the analysis of the preferences.
One parameter can reject a choice.

Michael: I agree.

For example: If a diamond looks brilliant from 99 out of 100 viewing directions and lighting circumstances, but has an obvious dark area from the face up viewing angle, it will be rejected as an Ideal.

Simple averaging, even using the cos square weighting, yields a high brilliance score for this rejected cut example.

Michael Cowing
www.acagemlab.com

 

[Garry H (Cut Nut)]

Hi Gents

For what it is worth - the Brilliancescope assumes equal light falls on a diamond from all angles. I prefer to see more weighting from 10 to 30 degrees as I did when establishing the data for HCA with the virtual Gilbertsonscope http://www.diamond-cut.com.au/07_gilbertson.htm

I prefer the stones with more red and green, and less pale blue. It is not an exact method - ut when you spread sheet it it takes on a pattern that is fairly reliable and does not have the error of assuming more light comes from shallower angles.