What's new
Photoshop Gurus Forum

Welcome to Photoshop Gurus forum. Register a free account today to become a member! It's completely free. Once signed in, you'll enjoy an ad-free experience and be able to participate on this site by adding your own topics and posts, as well as connect with other members through your own private inbox!

delta E


Ninanoki

Well-Known Member
Messages
89
Likes
0
i have a color sample in L,a,b mode used as 3d coordinates. I also know the coordinates of a few other colors (A1,A2,A3,etc) of which A1 and A2 are the closest. A2 in this case is the closest one. how should i measure how many percent of A2 color should I mix with A1 color to get the best match (lowest delta E)? are those dashed lines correct method?
xyz.jpg
A1 L:85 a:3 b:8
A2 L:80 a:5 b:13
Sample L:81 a:4 b:9
 
Last edited:
Hi Ninanoki
Its not really a Photoshop question yet here are some pointers.

- The 3 points in 3D space create a triangle
- You already know how to calculate the length of each side of the triangle (just ∆E equation between two points)
- This link gives you all needed formulas for triangles: http://www.ajdesigner.com/phptriangle/right_triangle_pythagorean_theorem_c.php
- In the scalene triangle section, given the length of the sides of the triangle you can calculate the angle between any two segments
- If either of the two angles between the two reference colors and your sample color is greater than 90 degrees, then the sample color is not between your reference colors
- The "height" or distance from the sample color to the line between the two reference colors is one of the equations given in the reference.
- From there, using the Pythagorean equation, you can use the calculated height and appropriate triangle side distance to calculate the distance from one of the reference colors to the ideal match point between the references
- The fraction of that distance from the reference color to the ideal color point, compared to the ∆E between the two reference colors is the same fractional point to pick between the L, a, and b components between the two reference colors to have the Lab coordinates of the closest match to your sample color point.

Those steps above could be combined into one single large equation yet thought stepping you through the geometry and providing the link to the equations would let you do the detailed work (i.e. teaching you to fish while not catching the fish for you).

Hope that does it for you.

PS - If you need a lot more you can always consider posting in the Fee for Service section: https://www.photoshopgurus.com/forum/photoshop-freelance-work-fee-for-service-section-/
 
Last edited:
this is my calculation results: red lines are important to me and they allow me to measure the percentage for colors to mix if theory is correct.

thebestcpu - please tell me only:
1)is this the correct line to measure the colors mixture percentage combination?

2)should i measure the "percentage combination" ex.70%A1+30%A2 for only two the closest to my sample colors or every combination like A1+A2,A1+A3,A1+B1...etc,A2+A3,A2+B2,etc

p.s.i made my shaders once again with lab mode and 16bits with all lightroom and ps settings you told be in other thread.thanks!
xyz2.jpg
 
is this physics? vector?
 
Hi Ninanoki

Just to reiterate, we are talking about an academic answer and not a practical answer. I continue to believe this biggest issue is being able to have accurate and stable numbers for your sample color and your reference colors. Until you have a workflow that has highly accurate sample and reference colors, the information below will not get you any closer to a correct shade.

1) Yes, assuming that you have some way in practice to accurately mix the two colors A1 and A2 (I don't know your mixing options) and also that the perpendicular point actually lies between A1 and A2 as opposed to an extension beyond those points.

2) I don't have access to the correct Lab values for the reference numbers so cannot give a complete answer with a quick forum response. You only need to include the line segments where the angle off the A1 to A2 line is less than 90 degrees. Practically you would not want to include any line segment between color references that are more than 1 off in either number or letter as mixing on a wider range no doubt introduces larger errors. With the remaining segments, pick the ideal color mix number that has the smallest ∆E to your sample color. Of course, if a J to a reference color is even closer, you would pick that reference color with no mixing. So academically, the aforementioned approach should give you the shortest color distance for any rationally mixed pair or reference colors.

---------------------

And for hybridmadness it is "analytic geometry" and very esoteric relative to the standard PS, LR, AI questions.

I think all the academic information needs to be tempered with limits in accuracy of measuring any given color and also that the human eye can barely distinguish a ∆E in color of 2.5 or more.
 
1. Guys, don't forget that the underlying problem is basically subtractive in nature, and not directly addressed in an additive space like Lab.

2. People have been arguing about the best algorithm to do this sort of thing since early in the last century and the discussion continues to the present. Google {kubelka monk paints} (hope I spelled it right) and {algorithm development paper app iPad}.

Once again, Nananoki, I don't want to seem disrespectful, but you don't even realize what u don't know even in just this one small area of the larger problem u have decided to take on. Your interest and perseverance are admirable, but u really need professional help unless you are just doing this as a hobby.

Tom
 
to get accurate sample (vita shaders) i'm going to use ministudio to eliminate shadows and highlights.
i must say that my first tooth crown (made with this calculations) was one of the best color matches in 6 years time so it absolutely works for 75% typical cases. 25% are cases were tooth structure is difficult, tooth has some characteristics etc and patient should see technician at his lab to match it directly. my technician also suggested not to use dental shaders but technical porcelain shades which he uses.

please just tell me how would You take a color sample from this picture (because necks are thick-more saturated and the edges are transparent)? this picure is not for my samples, just for a question.
1)should i wand select whole shader like A1 without white spots, filter-average and eyedrop?
2)or should i select fixed size rectangle in the middle of each shader?
sj.bdj.2012.144-f1.jpg
 
Last edited:
That's beyond my expertise yet there is lots of information available.

First article (abstract and full article for purchase) covers testing technique and some interesting conclusions on limitations of even the samples that are being tested.

The rest may also be of use.

http://www.ncbi.nlm.nih.gov/pubmed/17559535

http://webcache.googleusercontent.c.../PPR_VOL8_ISS3.ashx+&cd=6&hl=en&ct=clnk&gl=us

http://www3.unisi.it/dl2/20100304131719197/Corciolani.pdf.

The brochure in the next link is more generic information.

http://vitanorthamerica.com/wp-cont...entistry_Inside-Dental-Assisting_Paravina.pdf


The articles show the difficulties in achieving good numbers is quite difficult even with a much more accurate workflow than yours.

As a parallel. In a color manged system, getting the best match to a printers output is creating an ICC profile matched to the ink/printer/paper combination and have a photospectrometer measure the test "tiles" on the test print. If one cannot afford the photospectrometer, they can have the paper sent out to a lab to have it done for them. I suggest that if you have standard test chips that you want measure, that you have them sent out for high accuracy measurements. That is a one time cost to get more accurate numbers than you could with your mini-lab approach. You would still have the issues of accurate measurements or samples, yet at least your reference shades might have more accurate numbers this way. This is all hypothetical as even this is outside my expertise about dentistry and feasibility of having your own reference chips tested.

Since matching goes beyond just color and tone yet includes opalescence, luminescence and translucency I imagine anything that is being discussed in this thread and your previous threads is just scratching the surface. Best wishes on your project.
 
browsing those articles quickly right now i can see that L,a,b values have already been measured with photospectrometer for vita shades. i will check it with my values and see how precise it is. but going back to my last question...what part of the tooth should i eyedrop?even with photospectrometer.
 
Last edited:
Great links, John! :thumbsup: They are much more specific to the OP's field than the general articles on algorithms for simulation of mixing of real-world paints that I suggested in my previous post.

Best regards,

Tom
 
i heard many times on this forum that using photoshop is wrong for this kind of analysis because it's not precise. i keep that in mind but i don't want to buy Mac just to do more precise color drop. i just need to be a little more precise than human eye. especially men are not got (generally) in color analysis. so i need some help from excel and ps. i don't want to sell it or say it's perfect. SpectroShade is a pro tool (a pistol) that takes a picture of a tooth with some kind of light that can see deep inside tooth: it can see dentin under the transparent enamel layer. i can't do that by taking a simple photo of a tooth even with flash ring and macro lens. but even SpectroShade is not perfect. it tell me a color or best color mix with lowest dE but it doesn't say how thick such layer should be! it is not 3d way. so let me just a little more precise, not perfect;) but thanks for a great amount of help.

my current tooth color analysis project is based on two independent parts:
visual made in photoshop (even now it is better than human eye, but i'm still working on this)
mathematical made in excel: http://dentopolis.org/temp/quest.xlsx

excel math results:
1)in "Lab" sheet I enter color from photoshop color drop for a whole tooth, neck part,middle and bite part.
2)"LabAnalysis" sheet measures distance in 3d space from each shade to patients tooth color.
3)"Mieszanie Lab" uses this method to calculate mix of two closest colors: https://www.photoshopgurus.com/foru...1-lab-parameters-3d-space.html#post1533714119

i have a few questions:
1)should i use the same calculations for Lab and a,b separately? (sheets: ab, abAnalysis, Mieszanieab).L is responsible for lightness and i think it is important.
2)is my excel calculation a good way to find a best mix for two closest colors?
 
The nearest you will get in Photoshop is the script below, it compares the colour samlpler with your excel values, and gives you the results.
 

Attachments

i heard many times on this forum that using photoshop is wrong for this kind of analysis because it's not precise. i keep that in mind but i don't want to buy Mac just to do more precise color drop....

IMO, the slight numerical accuracy limitations of PS have almost nothing to do with the recommendations not to use PS. The real reason is that PS has none of the capabilities and features of a mathematical programming language, yet you are trying to use PS to perform image analysis, which is almost entirely mathematical in nature.

Although I may have missed this, I don't remember anyone recommending you switch from a PC to a Mac -- certainly, it wasn't me. IMO, one of your main external problems is your choice of software, not hardware platform.

Tom M
 
You should eyedrop the center of the tooth to get an average lab but to be more precise you can take 3 to 6 different measures
 

Back
Top