AVS Forum banner

1241 - 1260 of 1454 Posts

·
Registered
Joined
·
588 Posts
I'm still using my i1Pro2 until my jeti 1501 arrives, but I can confirm that the volumetric probe matching works ideally ones you get used to it.

With FCCM, it wasn't possible at all to achieve a profile, which is NIST conform. Especially with the red luminance. The difference between that color to my i1Pro2 was 2.5%- 3.5%. That's was the reason for that massive oversaturated look. So I decided to try the volumetric probe match, and I can honestly say that I am never going back and use FCCM. With the volumetric probe match, I don't have the red luminance error anymore, and the colors don't suffer from the oversaturated look.
Skintones are looking fantastic and very natural now. I can see a dramatic difference between the volumetric probe match 3D LUT calibrated pm vs. FCCM 3D LUT pm.

Attached, you see my verification of my K10-A volumetric profile vs. the i1Pro2 measurement.
I can highly recommend this great Excel sheet from Miki, which makes it super easy to verify the profile if someone interested in it sends me a pm.

 

·
Registered
Joined
·
861 Posts
That's great news for i1Pro2 users to look forward to.

For what it's worth, I did run a couple RGBCMY comparison runs between the i1Pro2 and the Jeti. The end results were consistent and repeatable.

Again, both the i1Pro2 and the Jeti were connected to read simultaneously (using ColourSpace) and set up to match FOV in a pitch black room. 3D LUT was reset to put the C9 panel in native Wide Gamut and OLED Light was set at 25 (approx 100 nits with 100% White pre-cal'd to AWP).



IMO, for the i1Pro2 to get this close to a meter costing 5 times its price is more than acceptable despite its larger difference in reading white.
 

·
Registered
Joined
·
9,174 Posts
That's great news for i1Pro2 users to look forward to.

For what it's worth, I did run a couple RGBCMY comparison runs between the i1Pro2 and the Jeti. The end results were consistent and repeatable.

Again, both the i1Pro2 and the Jeti were connected to read simultaneously (using ColourSpace) and set up to match FOV in a pitch black room. 3D LUT was reset to put the C9 panel in native Wide Gamut and OLED Light was set at 25 (approx 100 nits with 100% White pre-cal'd to AWP).



IMO, for the i1Pro2 to get this close to a meter costing 5 times its price is more than acceptable despite its larger difference in reading white.
I agree completely! Whoever's got that i1Pro2 now must be very pleased! ;)

I can't use the volumetric matching method with the software I have, but with the FCM Method, on my first attempt, I got a good match which was within the NIST limits (AFAICT with the aid of the calculator). I'm sure it's beginner's luck so don't feel the need to try it again!
 

·
Registered
Joined
·
79 Posts
I would have to disagree.
If the display is measuring the same xyY response, it will not have drifted.
And the stability/accuracy of any probe is enough to spot such a drift.
But, it is well know that the HVS has a real lack of stability, as it adapts to the general surround illuminant.

So we perceive a fixed colour temp differently if we have been spending time in the midday sun, compared to the golden hour.

Steve
Yeah I think your right, switching the color temperature from cold blue back to the perceptually matched one made it look the same again, there's probably some day to day vision fluctuations as well involved.
 

·
Registered
Joined
·
972 Posts
I did some comparison between the Four Colours Correction Matrix (FCCM), Three Matrixes Correction Method (TMC) and Multi-point Volumetric Probe Matching (MVPM) using Leon (liberator72) as a guinea pig (thanks mate!).

There it is:

TMC




FCCM




MVPM



Having ascertained that the method that returns results closest to the measurements of a spectroradiometer is the volumetric one, looking at the results of my tests, it seems clear to me that the second-ranked method is the good old FCCM, while TMC returns average results that are more encouraging but more distant from the truth.
 

·
Registered
Joined
·
79 Posts
@Anger.miki, @liberator72

Very interesting. Well, I too have developed a Three Matrix Correction processor using the stricter approach as defined in the LG Paper "Calibration of Colorimeters for RGBW Displays". I was wondering if you could share your data with me so that I can process it with my implementation and then you can maybe check if TMC scores any better.

Let me also explain what I mean by the stricter approach. Quoting from the foregoing paper:

"There are two possible ways of determining which sub-gamut contains a given measurement. The first is an iterative process with up to four steps that has near-perfect accuracy. The second is a single equation that is easier to implement...."


I have used the first method.

I will be needing the four patches measurements by both the spectro & the colorimeter, and the grey scale measurement data captured by the colorimeter without any correction applied. Thanks!
 

·
Registered
Joined
·
972 Posts
Sure, no problem. If I remember correctly, I used the first method too. I will give you via WhatsApp what I have:
- original profile done with FCCM correction;
- verification of the original profile (which has been corrected with TMC) with no correction applied;
- bdps
 
  • Like
Reactions: omarank

·
Registered
Joined
·
248 Posts
I did some comparison between the Four Colours Correction Matrix (FCCM), Three Matrixes Correction Method (TMC) and Multi-point Volumetric Probe Matching (MVPM) using Leon (liberator72) as a guinea pig (thanks mate!).


Having ascertained that the method that returns results closest to the measurements of a spectroradiometer is the volumetric one, looking at the results of my tests, it seems clear to me that the second-ranked method is the good old FCCM, while TMC returns average results that are more encouraging but more distant from the truth.

Am I looking at the graphs correctly or do I have it reversed? What meters were used?

-FCCM has more points measured under .5 dE than MVPM (975 to 874)?

-FCCM has 1 measurement over 1dE and its 1.06dE- barely over the threshold.
 

·
Registered
Joined
·
972 Posts
Klein K10A and Jeti 1501.

I'm sorry, I don't understand what is not clear to you.
 

·
Registered
Joined
·
4,291 Posts
I did some comparison between the Four Colours Correction Matrix (FCCM), Three Matrixes Correction Method (TMC) and Multi-point Volumetric Probe Matching (MVPM) using Leon (liberator72) as a guinea pig (thanks mate!).



There it is:



TMC









FCCM









MVPM







Having ascertained that the method that returns results closest to the measurements of a spectroradiometer is the volumetric one, looking at the results of my tests, it seems clear to me that the second-ranked method is the good old FCCM, while TMC returns average results that are more encouraging but more distant from the truth.


Deleted
 

·
Registered
Joined
·
79 Posts
Sure, no problem. If I remember correctly, I used the first method too. I will give you via WhatsApp what I have:
- original profile done with FCCM correction;
- verification of the original profile (which has been corrected with TMC) with no correction applied;
- bdps
Thanks, Miki!

As Miki found that the TMC processed data that I shared with him, after processing the verification data that he provided, matched perfectly with his processed data, it’s clear that he used the best/stricter implementation of TMC only when he did this comparison study. Also, as our individual processing results matched perfectly, we in a way implicitly QC’ed each other’s implementation.

There was one interesting observation. Using the two bpd files (the four color measurement data) for the spectro and the colorimeter respectively, I carried out the Meter Profiling Verification using both FCMM and TMC. The verification failed with FCMM (with a very small margin), but was successful with TMC with absolutely perfect matching. This seems to be in contrast to the comparison study, where FCMM was found to be overall more accurate than TMC. Miki and I will investigate this further as we have more data, and he will likely post more data on the comparison study for further analysis. (Screenshots of the two verifications are attached, first is for FCMM and second for TMC).
 

Attachments

·
Registered
Joined
·
2,200 Posts
When testing the LG Three Matrix method in comparison with FCMM and Volumetric matching, you should deliberately include colours that sit on the boundaries between the three different matrices.

Steve
 

·
Registered
Joined
·
861 Posts
There was one interesting observation. Using the two bpd files (the four color measurement data) for the spectro and the colorimeter respectively, I carried out the Meter Profiling Verification using both FCMM and TMC. The verification failed with FCMM (with a very small margin), but was successful with TMC with absolutely perfect matching. This seems to be in contrast to the comparison study, where FCMM was found to be overall more accurate than TMC.
I just want to clarify firstly that I was used purely for providing some measurement data here as I had just got new meters and was profiling anyway while testing ColourSpace. I can provide no real input into how this comparison has been made or what it attempts to show. As Miki said, I was just a guinea pig.

But I want to point out that at the time I performed the FCMM probe match, I checked it's outcome as I always do before beginning a profile, and there absolutely was not a green luminance error as shown in your screen shot there So I am curious as to how you have come to that outcome? I measured RGBW with the Jeti, followed by the Klein and finally with the corrected Klein and compared these immediately. If I recall correctly the largest error was in green luminance, but it was well below 1.0% and in the region of 0.3%.

For comparison, here is a screenshot of my FCMM created in Chromasurf and stored to the Klein internal memory a couple of days later (we know Klein does something different in it's FCMM implementation to LS/CS) and also a screenshot of the 64 point 4^3 cube MVPM match results. I do still have the BCS files from when I validated the FCMM match in ColourSpace too, but I have moved them from the original directory to clear some space so will have to find them, but they can and will be provided when I do which will confirm that there absolutely was not an error above 1.5% in green luminance.
 

Attachments

  • Like
Reactions: mrtickleuk

·
Registered
Joined
·
79 Posts
But I want to point out that at the time I performed the FCMM probe match, I checked it's outcome as I always do before beginning a profile, and there absolutely was not a green luminance error as shown in your screen shot there So I am curious as to how you have come to that outcome?
Through FCMM maths as described in the two FCMM papers. Using the two bpds (for the spectro and the colorimeter), I calculated the FCMM (shown in the screenshot). Then I processed the bpd data of the colorimeter with the FCMM to calculate the FCMM processed readings. And then the verification table was populated to see if the FCMM processed readings were within the NIST tolereance with the spectro readings.

Normally, when you select the active & the reference probe bpds in LS and then take the four color measurements, LS internally generates the FCMM and processes the readings with that matrix, as you probably already know. So, assuming there was no drift or variation in the readings of the colorimeter when taking raw measurements compared to when using FCMM, you would get exactly the same results as I got. Now as you said, your profiling verification was successful, that means the measurements by the colorimeter you got the second time were different from the first one, with a difference enough to keep the readings (after FCMM processing) within NIST tolerance.

To rule out any ambiguity, I think this FCMM, TMC and MVPM comparison exercise should be carried out again. We should also include some boundary colors in the profiling sequence as Steve suggested so that we can verify if we see similar results again.
 

Attachments

·
Registered
Joined
·
861 Posts
Through FCMM maths as described in the two FCMM papers. Using the two bpds (for the spectro and the colorimeter), I calculated the FCMM (shown in the screenshot). Then I processed the bpd data of the colorimeter with the FCMM to calculate the FCMM processed readings. And then the verification table was populated to see if the FCMM processed readings were within the NIST tolereance with the spectro readings.

Normally, when you select the active & the reference probe bpds in LS and then take the four color measurements, LS internally generates the FCMM and processes the readings with that matrix, as you probably already know. So, assuming there was no drift or variation in the readings of the colorimeter when taking raw measurements compared to when using FCMM, you would get exactly the same results as I got. Now as you said, your profiling verification was successful, that means the measurements by the colorimeter you got the second time were different from the first one, with a difference enough to keep the readings (after FCMM processing) within NIST tolerance.

To rule out any ambiguity, I think this FCMM, TMC and MVPM comparison exercise should be carried out again. We should also include some boundary colors in the profiling sequence as Steve suggested so that we can verify if we see similar results again.
Thank you for you explanation. Yes, I do understand the methods to creating the matrix using the RAW readings from the BPD files, this is something I've done many times in the past in order to import the same correction used in LS to CalMAN. I was just curious as to how it came to show the green luminance error when in reality when I validated the FCMM in real time there was no such error.

But anyway, it's semantics at this point, and as stated my measurement data was used purely because I was profiling that day as I had new probes, and was experimenting with a new patchset (kindly provided by @BlackJoker, one picture mode with FCMM and one with MVPM) and happened to mention to Miki that I was curious about the new TCM method.

I'm happy to take some further measurements if required.
 

·
Registered
Joined
·
9,174 Posts
Thanks, Miki!

As Miki found that the TMC processed data that I shared with him, after processing the verification data that he provided, matched perfectly with his processed data, it’s clear that he used the best/stricter implementation of TMC only when he did this comparison study. Also, as our individual processing results matched perfectly, we in a way implicitly QC’ed each other’s implementation.

There was one interesting observation. Using the two bpd files (the four color measurement data) for the spectro and the colorimeter respectively, I carried out the Meter Profiling Verification using both FCMM and TMC. The verification failed with FCMM (with a very small margin), but was successful with TMC with absolutely perfect matching. This seems to be in contrast to the comparison study, where FCMM was found to be overall more accurate than TMC. Miki and I will investigate this further as we have more data, and he will likely post more data on the comparison study for further analysis. (Screenshots of the two verifications are attached, first is for FCMM and second for TMC).
Thankyou both for doing this, and to @liberator72 for his hard work and patience taking all those readings.

Can I ask about the "Profiling Verification" please. I understand that the goal is to see if the xy are +/-0.0001 and the Y is +/- 1.5%, but can't find a reference for those thresholds. Also, if those are the rules for the FCM method, how did you decide what rules to use for the TCM method?
 

·
Registered
Joined
·
17,806 Posts
Thankyou both for doing this, and to @liberator72 for his hard work and patience taking all those readings.

Can I ask about the "Profiling Verification" please. I understand that the goal is to see if the xy are +/-0.0001 and the Y is +/- 1.5%, but can't find a reference for those thresholds. Also, if those are the rules for the FCM method, how did you decide what rules to use for the TCM method?
+/-0.0015, not 0.0001 :)
 
1241 - 1260 of 1454 Posts
Top