Optical analysis of mounted diamonds

[denverappraiser]

Garry, Sergey, Marty etc.

Shouldn’t it be possible to use a software solution to convert a table up photograph of a diamond into a DiamCalc or .srn style model of it? It may be necessary to use an Okuda or Gemprint style system for aligning the stone so that photo is taken from a position that is directly perpendicular to the table but this doesn’t seems like that serious a restriction. It seems like it would be reasonably possible to use software to recognize facet junctions and other key attributes of the stone. Armed with this model, couldn’t you make a decent estimation of the less obvious dimensions, like crown and pavilion angle on mounted stones?

Neil Beaty
Independent Appraisals in Denver

 

[jackson]

That''s kind of what I was asking.....How do you do measurements of a stone in a mounting/setting?

 

[strmrdr]

im going to answer and see how my answers line up with the experts.

Angles no.
girdle no so cant duplicate a .srn file.
pavilion info would be extremely limited.
I doubt if you could extract enough information to be useful overall.

 

[Serg]

Date: 12/12/2004 9:38:12 AM
Author:denverappraiser

Garry, Sergey, Marty etc.

Shouldn’t it be possible to use a software solution to convert a table up photograph of a diamond into a DiamCalc or .srn style model of it? It may be necessary to use an Okuda or Gemprint style system for aligning the stone so that photo is taken from a position that is directly perpendicular to the table but this doesn’t seems like that serious a restriction. It seems like it would be reasonably possible to use software to recognize facet junctions and other key attributes of the stone. Armed with this model, couldn’t you make a decent estimation of the less obvious dimensions, like crown and pavilion angle on mounted stones?

Neil Beaty
Independent Appraisals in Denver

Hi, Neil

It is possible for symmetrical diamond if great accuracy is not necessary.
publish photo of your diamond and I will show result.

Do you know diameter and height?

 

[Superidealist]

I think what Neil is talking about is beginning with the face up appearance and working backwards to get the dimensions. That sounds like a much more difficult proposition than the reverse.

 

[strmrdr]

Here is a 10x photo of a diamond from whiteflash.com that Iv saved the sarin and idealscope image for.

6.53- 6.55 - 3.98

 

[denverappraiser]

Storm, et. al.

Crown angles: Back in the dark ages of the 70's, the way GIA taught us to esitmate crown and pavillion angles was by comparing the image of the pavillion main as seen through the table with the image as refracted through the crown. This seems like a task that a computer would be well suited for.

Pavillion angles: The same trick can be done using the reflection of the table in the pavillion mains.

Table: easy

Facet allignment, yaw, centering of the table and culet, out-of-round etc. It's all there.

Girdle. At least one dimension would be required in order to generate a scale. Diameters and depth are usually easy to directly measure and can be data entered and this is regularly known information anyway because of a lab report. With the above items caluclated, it should be possible to make a pretty decent guess at the girdle but I agree that this is a bit of a stretch. As with the basic dimensions, most lab reports contain a mention of the girdle. If the photo is good enough, you can actually see the reflection of the girdle in the pavillion facets by the way.

I'm looking for a way to get the data that is omitted from the lab report without unmounting the stone. It seems like some of it is inculcated into the photo if we were just prepared to look with the right kind of eyes.

Neil Beaty
Independent Appraisals in Denver

 

[strmrdr]

denverappraiser,
Thanks for the info, interesting.
Now im even more curious to see what serg comes up with :}

 

[denverappraiser]

Sergey,

I don''t have a diamond in mind but the one posted by Storm will be a good example to play with. I wasn''t really expecting great accuracy but I don''t actually see why you can''t get it if the photo is sufficiently detailed.

Neil

 

[Serg]

Date: 12/12/2004 11

5:31 AM
Author: strmrdr
im going to answer and see how my answers line up with the experts.

Angles no.
girdle no so cant duplicate a .srn file.
pavilion info would be extremely limited.
I doubt if you could extract enough information to be useful overall.

Hi, Strmrdr

See example of good old virtual work, Rhino sent photo only.

see result.

http://www.diamondtalk.com/forums/showthread.php?t=17342&perpage=25&highlight=serg&pagenumber=8
http://www.diamondtalk.com/forums/showthread.php?t=17342&perpage=25&highlight=serg&pagenumber=9

 

[strmrdr]

Thanks serg,
Im going to study that thread.
It will be interesting to see how close you get on the one above.

 

[denverappraiser]

Am I correct that the alignment of the table when the photo is taken is important? I can''t think of any way to adjust for a variation here. It also seems possible that the distance from the camera to the stone when the photo is taken might be important as well. If we are comparing two different images of the pavilion mains and drawing conclusions from that, it might be relevant that the direction of viewing is slightly different. The best would be a photo taken with a camera that is relatively far away.

Neil Beaty
Independent Appraisals in Denver

 

[Serg]

Date: 12/12/2004 11:27:47 AM
Author: strmrdr
Here is a 10x photo of a diamond from whiteflash.com that Iv saved the sarin and idealscope image for.

6.53- 6.55 - 3.98

Very bad photo for such work

I am prefer Rhino Photos.
But
m=1.01(1.02) ct
PavAngle 40.75 ( Pav min 40.70 pav max 40.
CrownAngle= 35.8( very strange. may be azimuth of crown girdle is not classical)
Girdle : Very Thin
LOwer facet =79%
Star facet =53%
Table=55%

 

[Serg]

Date: 12/12/2004 12

4:28 PM
Author: denverappraiser

Am I correct that the alignment of the table when the photo is taken is important? I can''t think of any way to adjust for a variation here. It also seems possible that the distance from the camera to the stone when the photo is taken might be important as well. If we are comparing two different images of the pavilion mains and drawing conclusions from that, it might be relevant that the direction of viewing is slightly different. The best would be a photo taken with a camera that is relatively far away.

Neil Beaty
Independent Appraisals in Denver

Yes, Yes

 

[Serg]

.

 

[denverappraiser]

Serg,

Thank you. That''s just exactly what I''m asking about. You''re amazing. How did you get the wireframe diagram from the picture?

You gave very precise answers but you said before that your accuracy would not be very good. Can you explain? Also, you mentioned that Rhino photos are more appropriate for this kind of work. Why?

Neil

 

[strmrdr]

sarin for it...

 

[strmrdr]

is image...

 

[strmrdr]

ags cert

 

[strmrdr]

My conclusion with this image the calculations while interesting arent accurate enough to be useful.
The crown angle being way off was the biggest variation.
Thanks serg.

 

[strmrdr]

This image any better?
3.97x3.99x2.42

 

[denverappraiser]

Storm,

Actually, I''m quite impressed with how well he did. He missed the crown by 1 degree, the table is dead on, the pavillion angle is dead on and he even picked up the thin girdle. Since he had no control over the photograph itself, I''ld call his results spectacular.

Neil

 

[strmrdr]

Date: 12/12/2004 1:35:33 PM
Author: denverappraiser
Storm,


Actually, I''m quite impressed with how well he did. He missed the crown by 1 degree, the table is dead on, the pavillion angle is dead on and he even picked up the thin girdle. Since he had no control over the photograph itself, I''ld call his results spectacular.


Neil

ok...
But to be usefull it needs to be close enough that you can use the other tools to judge the cut.
Run both sets of numbers thru the hca and look at the results.
Its a no go,,,,

 

[Garry H (Cut Nut)]

Neil - Richard Sherwood has been doing this for some time - but I think he matches the ideal-scope image with a DiamCalc Ideal-scope view.

You can also do this from a good photo.

 

[denverappraiser]

Serg,

If you could take the photograph under controlled circumstances, do you think this technique can produce results that would be useful with the HCA? Sarin data can only be trusted to, at best, 0.2 deg. so you''re only off by a factor of 5 and we have no idea how much of that is attributed to the photography. The more I think about this, the more impressed I am with your work.

Neil

 

[Garry H (Cut Nut)]

Also neil - the Liddicoat Rapid Sight Pavilion Depth Estimation Method is flawed.
These stones all have the exact same pavilion depth - and you would get different results - agreed?

i have published a review of this in the Australian Gemmologst.

 

[denverappraiser]

Garry,

I've not found the sight estimation method to be particularly useful but it's an interesting concept. I haven’t thought about it for 20 years and it’s a little hard to discuss it meaningfully. There are obviously a fair number of variables involved but it seems like the data is all there and that it's just a matter of correctly analyzing what is presented. I don’t claim to be anywhere near as smart as Sergey (or you or Richard for that matter) but I don’t really understand why this system isn’t in common use. This is a computers job if ever there was one. Sergey just demonstrated that it works pretty well and the endorsement from Richard is an awfully good sign. The three examples that you present are all about the crown geometry. This is a little difficult to do in your head but it doesn't seem like it makes the whole idea useless. In playing with the DiamCalc software, this is sort of a time consuming process but it seems like something that would automate well, especially if the photograph comes from standardized conditions.


I don’t have the pleasure of subscribing to Australian Gemmologist so I haven’t had the opportunity to read your thoughts there. Are they still teaching this at GIA?

Neil

 

[Serg]

Date: 12/12/2004 1:20:47 PM
Author: strmrdr
is image...

This photo is better fro angles, bit worse for table

Pav 40.7

Cr 34.5

table 56

GirdleValley 1.2%

Mass 1.027

Lowerfacet 80%

Star facet 53%

 

[JohnQuixote]

Very impressive, Serg.

Strm, you're a hard man, but you have a point - it will need to be more accurate to be accepted as practical.

That photo was set up for aesthetics, not for analysis.
The crown issues may have to do with the setup (just guessing).

Here is another photo which would probably work better.
Serg - will this suit your purposes?

I don't have stats at my fingertips on a Sunday.
I'll need to track down the Sarin/measurements on it tomorrow.

 

[strmrdr]

kewl serg
It looks like the crown angle is the hardest to nail down.
Is that currect?
Would controlled and known conditions help nail the crown angle down?