Physics Talk :: View topic - Accurate formula to measure display GAMMA !

Physics Talk
Physics Talk

FAQ

Memberlist

Usergroups

Profile

Education Forums

Accurate formula to measure display GAMMA !
Goto page 1, 2  Next

       Physics Talk Forum Index -> Color and Light

View previous topic :: View next topic

Author Message

Rodney
Guest

Posted: Thu Jun 12, 2008 12:30 pm   Post subject: Accurate formula to measure display GAMMA !

Dear ALL,

I am in process of measuring GAMMA of my display device, and wanted
some advice for the same. I refer http://www.brucelindbloom.com/index.html?Eqn_BestGamma.html
to calculate the gamma value, but not getting the correct GAMMA value
as expected.

Steps, i used to measure GAMMA:

a) Measured Luminance for various grayscale using a calibrator.
GrayScale values (val, val, val)
where val range from 1 to 254

b) Normalized grayscale array by dividing each value by 255

c) Normalized luminance array by dividing each value by highest value
of luminance ( 197 in my case)

d) Used formula mentioned in http://www.brucelindbloom.com/index.html?Eqn_BestGamma.html
to compute GAMMA

Also plotted graph on excel for those values. The graph is a proper
curve as expected. BUT, the gamma comes out to be 2.01 for a display
device which has a GAMMA of 2.2. I verfied the GAMMA of device by
measuring it with a third party application which gives GAMMA as 2.21.

Also i tested with other gamma values, my application could not give
correct GAMMA values.

What can be the issue ?
Is the formula accurate mentioned in this site ?
Can you please suggest some alternative formulae?

Please comment.

Thanks,
Rodney

Back to top

Ads

Advertising
Sponsor

Gernot Hoffmann
Guest

Posted: Thu Jun 12, 2008 12:48 pm   Post subject: Re: Accurate formula to measure display GAMMA !

Rodney schrieb:

Quote:

Dear ALL,

I am in process of measuring GAMMA of my display device, and wanted
some advice for the same. I refer http://www.brucelindbloom.com/index.html?Eqn_BestGamma.html
to calculate the gamma value, but not getting the correct GAMMA value
as expected.

Steps, i used to measure GAMMA:

a) Measured Luminance for various grayscale using a calibrator.
GrayScale values (val, val, val)
where val range from 1 to 254

b) Normalized grayscale array by dividing each value by 255

c) Normalized luminance array by dividing each value by highest value
of luminance ( 197 in my case)

d) Used formula mentioned in http://www.brucelindbloom.com/index.html?Eqn_BestGamma.html
to compute GAMMA

Also plotted graph on excel for those values. The graph is a proper
curve as expected. BUT, the gamma comes out to be 2.01 for a display
device which has a GAMMA of 2.2. I verfied the GAMMA of device by
measuring it with a third party application which gives GAMMA as 2.21.

Also i tested with other gamma values, my application could not give
correct GAMMA values.

What can be the issue ?
Is the formula accurate mentioned in this site ?
Can you please suggest some alternative formulae?

Please comment.

Thanks,
Rodney

Rodney,

IMHO the mathematical formulation as used by B.L. is too
simple. Alternatives are here:

http://www.fho-emden.de/~hoffmann/measgamma10022004.pdf

The formulation by logarithms lead in my tests to very bad
results.

Best regards --Gernot Hoffmann

Back to top

Ads

Advertising
Sponsor

Graeme Gill
Guest

Posted: Thu Jun 12, 2008 6:32 pm   Post subject: Re: Accurate formula to measure display GAMMA !

Rodney wrote:

Quote:

d) Used formula mentioned in http://www.brucelindbloom.com/index.html?Eqn_BestGamma.html
to compute GAMMA

Also plotted graph on excel for those values. The graph is a proper
curve as expected. BUT, the gamma comes out to be 2.01 for a display
device which has a GAMMA of 2.2. I verfied the GAMMA of device by
measuring it with a third party application which gives GAMMA as 2.21.

What makes the third part application your golden reference ?

Given that a real device will never have a perfect power curve,
there are an infinite number of ways it can be approximated by a pure
power curve.
Methods off the top of my head :

Compute the 50% input value ratio to the 100% value. Call the
gamma the power that would give this same ratio.

Same as above, but subtract the output value black offset.

Same as above, but remove the non-zero black value as a device offset.

Compute the least squares best fit of a gamma curve measured in device output space.

Compute the least squares best fit of a gamma curve measured in L*a*b* space.

Compute the least maximum error fit ......

etc.

Take your pick.

Graeme Gill.

Back to top

Ads

Advertising
Sponsor

Rodney
Guest

Posted: Fri Jun 13, 2008 12:19 am   Post subject: Re: Accurate formula to measure display GAMMA !

On Jun 12, 9:48 pm, Gernot Hoffmann <hoffm...@fho-emden.de> wrote:

Quote:

Rodney schrieb:

Dear ALL,

I am in process of measuring GAMMA of my display device, and wanted
some advice for the same. I referhttp://www.brucelindbloom.com/index.html?Eqn_BestGamma.html
to calculate the gamma value, but not getting the correct GAMMA value
as expected.

Steps, i used to measure GAMMA:

a) Measured Luminance for various grayscale using a calibrator.
GrayScale values (val, val, val)
where val range from 1 to 254

b) Normalized grayscale array by dividing each value by 255

c) Normalized luminance array by dividing each value by highest value
of luminance ( 197 in my case)

d) Used formula mentioned inhttp://www.brucelindbloom.com/index.html?Eqn_BestGamma.html
to compute GAMMA

Also plotted graph on excel for those values. The graph is a proper
curve as expected. BUT, the gamma comes out to be 2.01 for a display
device which has a GAMMA of 2.2. I verfied the GAMMA of device by
measuring it with a third party application which gives GAMMA as 2.21.

Also i tested with other gamma values, my application could not give
correct GAMMA values.

What can be the issue ?
Is the formula accurate mentioned in this site ?
Can you please suggest some alternative formulae?

Please comment.

Thanks,
Rodney

Rodney,

IMHO the mathematical formulation as used by B.L. is too
simple. Alternatives are here:

http://www.fho-emden.de/~hoffmann/measgamma10022004.pdf

The formulation by logarithms lead in my tests to very bad
results.

Best regards --Gernot Hoffmann- Hide quoted text -

- Show quoted text -

HI
Below is data, can anybody compute the gamma from his best/approx best
algo and share the algo? I get 2.051916 gamma using
inhttp://www.brucelindbloom.com/index.html?Eqn_BestGamma.html
//////////
Input_stimulus
0.003922
0.023529
0.043137
0.062745
0.082353
0.101961
0.121569
0.141176
0.160784
0.180392
0.2
0.219608
0.239216
0.258824
0.278431
0.298039
0.317647
0.337255
0.356863
0.376471
0.396078
0.415686
0.435294
0.454902
0.47451
0.494118
0.513725
0.533333
0.552941
0.572549
0.592157
0.611765
0.631373
0.65098
0.670588
0.690196
0.709804
0.729412
0.74902
0.768627
0.788235
0.807843
0.827451
0.847059
0.866667
0.886275
0.905882
0.92549
0.945098
0.964706
0.984314

//////////
Luma
0.001044
0.003034
0.004183
0.005075
0.006317
0.007405
0.0088
0.010078
0.012046
0.014251
0.016191
0.019073
0.022423
0.026112
0.029536
0.033975
0.039181
0.046165
0.052021
0.058742
0.066459
0.075504
0.08405
0.09491
0.10624
0.116753
0.129794
0.143859
0.158203
0.173312
0.190866
0.208449
0.23089
0.251784
0.273881
0.301429
0.32794
0.359782
0.389697
0.425495
0.461436
0.502597
0.547773
0.593995
0.644025
0.702238
0.755649
0.81671
0.867771
0.926902
1
//////////
Rodney.

Back to top

Ads

Advertising
Sponsor

Gernot Hoffmann
Guest

Posted: Fri Jun 13, 2008 7:23 am   Post subject: Re: Accurate formula to measure display GAMMA !

Rodney schrieb:

Quote:

On Jun 12, 9:48ï¿½pm, Gernot Hoffmann <hoffm...@fho-emden.de> wrote:
Rodney schrieb:

Dear ALL,

I am in process of measuring GAMMA of my display device, and wanted
some advice for the same. ï¿½I referhttp://www.brucelindbloom.com/index.html?Eqn_BestGamma.html
to calculate the gamma value, but not getting the correct GAMMA value
as expected.

Steps, i used to measure GAMMA:

a) Measured Luminance ï¿½for various grayscale using ï¿½a calibrator.
GrayScale values (val, val, val)
where val range from 1 to 254

b) Normalized grayscale array by dividing each value by 255

c) Normalized luminance array by dividing each value by highest value
of luminance ( 197 in my case)

d) Used formula mentioned inhttp://www.brucelindbloom.com/index.html?Eqn_BestGamma.html
to compute GAMMA

Also plotted graph on excel for those values. The graph is a proper
curve as expected. BUT, the gamma comes out to be 2.01 for a display
device which has a GAMMA of 2.2. I verfied the GAMMA of device by
measuring it with a third party application which gives GAMMA as 2.21.

Also i tested with other gamma values, my application could not give
correct GAMMA values.

What can be the issue ?
Is the formula accurate mentioned in this site ?
Can you please suggest some alternative formulae?

Please comment.

Thanks,
Rodney

Rodney,

IMHO the mathematical formulation as used by B.L. is too
simple. Alternatives are here:

http://www.fho-emden.de/~hoffmann/measgamma10022004.pdf

The formulation by logarithms lead in my tests to very bad
results.

Best regards --Gernot Hoffmann- Hide quoted text -

- Show quoted text -

HI
Below is data, can anybody compute the gamma from his best/approx best
algo and share the algo? I get 2.051916 gamma using
inhttp://www.brucelindbloom.com/index.html?Eqn_BestGamma.html
//////////
Input_stimulus
0.003922
0.023529
0.043137
0.062745
0.082353
0.101961
0.121569
0.141176
0.160784
0.180392
0.2
0.219608
0.239216
0.258824
0.278431
0.298039
0.317647
0.337255
0.356863
0.376471
0.396078
0.415686
0.435294
0.454902
0.47451
0.494118
0.513725
0.533333
0.552941
0.572549
0.592157
0.611765
0.631373
0.65098
0.670588
0.690196
0.709804
0.729412
0.74902
0.768627
0.788235
0.807843
0.827451
0.847059
0.866667
0.886275
0.905882
0.92549
0.945098
0.964706
0.984314

//////////
Luma
0.001044
0.003034
0.004183
0.005075
0.006317
0.007405
0.0088
0.010078
0.012046
0.014251
0.016191
0.019073
0.022423
0.026112
0.029536
0.033975
0.039181
0.046165
0.052021
0.058742
0.066459
0.075504
0.08405
0.09491
0.10624
0.116753
0.129794
0.143859
0.158203
0.173312
0.190866
0.208449
0.23089
0.251784
0.273881
0.301429
0.32794
0.359782
0.389697
0.425495
0.461436
0.502597
0.547773
0.593995
0.644025
0.702238
0.755649
0.81671
0.867771
0.926902
1
//////////
Rodney.

Rodney,

I had plugged your data into my program, version B.
The result is Gamma=3.09. This data set isn't even near to 2.0,
but much nearer to 3.0, as it can be shown by drawing such a
function without offset.
Your function table should be built by input 0 (black) as first
input and input 1 (white) as last input. This is necessary for
the normalization. Output should be the luminance in cd/m2.
The normalization can be done by the program.
Please use only 1+ 8 value pairs (input 0.0, 0.125, ..., 1.0).
If you prepare a new set, then I can calculate gamma again.
Meanwhile you may check gamma for R,G,B and for gray by
test patterns, mainly p.7:
http://www.fho-emden.de/~hoffmann/caltutor270900.pdf

Best regards --Gernot Hoffmann

Back to top

Ads

Advertising
Sponsor

Rodney
Guest

Posted: Fri Jun 13, 2008 8:01 am   Post subject: Re: Accurate formula to measure display GAMMA !

Dear Gernot,
Please find below the updated table, Please share your views for the
same
//////////
Input Stimulus
0
0.12549
0.25098
0.376471
0.501961
0.627451
0.752941
0.878431
1
//////////
Output Stimulus
0.00092
0.002908
0.019715
0.089258
0.203187
0.353551
0.532005
0.744232
1
/////////

Also what approach are you following ?
Rodney

Gernot Hoffmann wrote:

Quote:

Rodney schrieb:
On Jun 12, 9:48ï¿½pm, Gernot Hoffmann <hoffm...@fho-emden.de> wrote:
Rodney schrieb:

Dear ALL,

I am in process of measuring GAMMA of my display device, and wanted
some advice for the same. ï¿½I referhttp://www.brucelindbloom.com/index.html?Eqn_BestGamma.html
to calculate the gamma value, but not getting the correct GAMMA value
as expected.

Steps, i used to measure GAMMA:

a) Measured Luminance ï¿½for various grayscale using ï¿½a calibrator.
GrayScale values (val, val, val)
where val range from 1 to 254

b) Normalized grayscale array by dividing each value by 255

c) Normalized luminance array by dividing each value by highest value
of luminance ( 197 in my case)

d) Used formula mentioned inhttp://www.brucelindbloom.com/index.html?Eqn_BestGamma.html
to compute GAMMA

Also plotted graph on excel for those values. The graph is a proper
curve as expected. BUT, the gamma comes out to be 2.01 for a display
device which has a GAMMA of 2.2. I verfied the GAMMA of device by
measuring it with a third party application which gives GAMMA as 2.21.

Also i tested with other gamma values, my application could not give
correct GAMMA values.

What can be the issue ?
Is the formula accurate mentioned in this site ?
Can you please suggest some alternative formulae?

Please comment.

Thanks,
Rodney

Rodney,

IMHO the mathematical formulation as used by B.L. is too
simple. Alternatives are here:

http://www.fho-emden.de/~hoffmann/measgamma10022004.pdf

The formulation by logarithms lead in my tests to very bad
results.

Best regards --Gernot Hoffmann- Hide quoted text -

- Show quoted text -

HI
Below is data, can anybody compute the gamma from his best/approx best
algo and share the algo? I get 2.051916 gamma using
inhttp://www.brucelindbloom.com/index.html?Eqn_BestGamma.html
//////////
Input_stimulus
0.003922
0.023529
0.043137
0.062745
0.082353
0.101961
0.121569
0.141176
0.160784
0.180392
0.2
0.219608
0.239216
0.258824
0.278431
0.298039
0.317647
0.337255
0.356863
0.376471
0.396078
0.415686
0.435294
0.454902
0.47451
0.494118
0.513725
0.533333
0.552941
0.572549
0.592157
0.611765
0.631373
0.65098
0.670588
0.690196
0.709804
0.729412
0.74902
0.768627
0.788235
0.807843
0.827451
0.847059
0.866667
0.886275
0.905882
0.92549
0.945098
0.964706
0.984314

//////////
Luma
0.001044
0.003034
0.004183
0.005075
0.006317
0.007405
0.0088
0.010078
0.012046
0.014251
0.016191
0.019073
0.022423
0.026112
0.029536
0.033975
0.039181
0.046165
0.052021
0.058742
0.066459
0.075504
0.08405
0.09491
0.10624
0.116753
0.129794
0.143859
0.158203
0.173312
0.190866
0.208449
0.23089
0.251784
0.273881
0.301429
0.32794
0.359782
0.389697
0.425495
0.461436
0.502597
0.547773
0.593995
0.644025
0.702238
0.755649
0.81671
0.867771
0.926902
1
//////////
Rodney.

Rodney,

I had plugged your data into my program, version B.
The result is Gamma=3.09. This data set isn't even near to 2.0,
but much nearer to 3.0, as it can be shown by drawing such a
function without offset.
Your function table should be built by input 0 (black) as first
input and input 1 (white) as last input. This is necessary for
the normalization. Output should be the luminance in cd/m2.
The normalization can be done by the program.
Please use only 1+ 8 value pairs (input 0.0, 0.125, ..., 1.0).
If you prepare a new set, then I can calculate gamma again.
Meanwhile you may check gamma for R,G,B and for gray by
test patterns, mainly p.7:
http://www.fho-emden.de/~hoffmann/caltutor270900.pdf

Best regards --Gernot Hoffmann

Back to top

Ads

Advertising
Sponsor

Gernot Hoffmann
Guest

Posted: Fri Jun 13, 2008 10:41 am   Post subject: Re: Accurate formula to measure display GAMMA !

Rodney schrieb:

Quote:

Dear Gernot,
Please find below the updated table, Please share your views for the
same
//////////
Input Stimulus
0
0.12549
0.25098
0.376471
0.501961
0.627451
0.752941
0.878431
1
//////////
Output Stimulus
0.00092
0.002908
0.019715
0.089258
0.203187
0.353551
0.532005
0.744232
1
/////////

Also what approach are you following ?
Rodney

Rodney,

it's the mathematical model Type B here on p.3:
http://www.fho-emden.de/~hoffmann/measgamma10022004.pdf
The curve fitting is based on numerical optimization (Steepest
Descent), as explained.

Your actual data deliver Gamma=2.175 and Offset=-0.023.
This says: your monitor is too dark at the dark end. Dark
input values will be clipped visually.

Assumed, your white has 100cd/m2, then your black would be
about 0.1cd/m2. More realistic is 0.3cd/m2.

Feel free to send new measured data after a re-calibration
concerning the blackpoint.

Best regards --Gernot Hoffmann

Back to top

Ads

Advertising
Sponsor

Rodney
Guest

Posted: Mon Jun 16, 2008 1:20 am   Post subject: Re: Accurate formula to measure display GAMMA !

On Jun 13, 7:41 pm, Gernot Hoffmann <hoffm...@fho-emden.de> wrote:

Quote:

Rodney schrieb:

Dear Gernot,
Please find below the updated table, Please share your views for the
same
//////////
Input Stimulus
0
0.12549
0.25098
0.376471
0.501961
0.627451
0.752941
0.878431
1
//////////
Output Stimulus
0.00092
0.002908
0.019715
0.089258
0.203187
0.353551
0.532005
0.744232
1
/////////

Also what approach are you following ?
Rodney

Rodney,

it's the mathematical model Type B here on p.3:http://www.fho-emden.de/~hoffmann/measgamma10022004.pdf
The curve fitting is based on numerical optimization (Steepest
Descent), as explained.

Your actual data deliver Gamma=2.175 and Offset=-0.023.
This says: your monitor is too dark at the dark end. Dark
input values will be clipped visually.

Assumed, your white has 100cd/m2, then your black would be
about 0.1cd/m2. More realistic is 0.3cd/m2.

Feel free to send new measured data after a re-calibration
concerning the blackpoint.

Best regards --Gernot Hoffmann- Hide quoted text -

- Show quoted text -

Dear Gernot,
Below is the monitor current measurement
Brightness: 213 cd/m2
Black level: 0.2 cd/m2
CT: 6500
///////
Input
0.000000
0.125000
0.250000
0.375000
0.500000
0.625000
0.750000
0.875000
0.996094
////
Output
0.000857
0.004562
0.026489
0.088656
0.183950
0.326179
0.518590
0.744004
1.000000
////
B/w I hope you are using "single dimension steepest descent" algo as u
mentioned in the previous mail. Can you share xls sheet implementing
the same? Or C/C++ implementation?
Thanks,
Rodney

Back to top

Ads

Advertising
Sponsor

Gernot Hoffmann
Guest

Posted: Mon Jun 16, 2008 6:48 am   Post subject: Re: Accurate formula to measure display GAMMA !

Rodney schrieb:

Quote:

On Jun 13, 7:41ï¿½pm, Gernot Hoffmann <hoffm...@fho-emden.de> wrote:
Rodney schrieb:

Dear Gernot,
Please find below the updated table, Please share your views for the
same
//////////
Input Stimulus
0
0.12549
0.25098
0.376471
0.501961
0.627451
0.752941
0.878431
1
//////////
Output Stimulus
0.00092
0.002908
0.019715
0.089258
0.203187
0.353551
0.532005
0.744232
1
/////////

Also what approach are you following ?
Rodney

Rodney,

it's the mathematical model Type B here on p.3:http://www.fho-emden.de/~hoffmann/measgamma10022004.pdf
The curve fitting is based on numerical optimization (Steepest
Descent), as explained.

Your actual data deliver Gamma=2.175 and Offset=-0.023.
This says: your monitor is too dark at the dark end. Dark
input values will be clipped visually.

Assumed, your white has 100cd/m2, then your black would be
about 0.1cd/m2. More realistic is 0.3cd/m2.

Feel free to send new measured data after a re-calibration
concerning the blackpoint.

Best regards --Gernot Hoffmann- Hide quoted text -

- Show quoted text -

Dear Gernot,
Below is the monitor current measurement
Brightness: 213 cd/m2
Black level: 0.2 cd/m2
CT: 6500
///////
Input
0.000000
0.125000
0.250000
0.375000
0.500000
0.625000
0.750000
0.875000
0.996094
////
Output
0.000857
0.004562
0.026489
0.088656
0.183950
0.326179
0.518590
0.744004
1.000000
////
B/w I hope you are using "single dimension steepest descent" algo as u
mentioned in the previous mail. Can you share xls sheet implementing
the same? Or C/C++ implementation?
Thanks,
Rodney

Rodney,

I'm using the algorithm chapter 4.2,Steepest descent with one-dimen-
sional search.
C/C++ is not a available, only the algorithm and PostScript code.
If you have Photoshop or InDesign or Illustrator or Ghostscript, then
you can use my Postscript code Reference [5] (rename as EPS).
Just replace the one table GamWht for White (gray) by your data, using
e.g. WordPad (plain text). Input left, output right.

Here is the result for your recent data:
http://www.fho-emden.de/~hoffmann/rodney16062008.gif

Note that my program ignores the first value (at 0.0), because this is
mostly uncertain. The next value is IMHO still too dark which leads
in
the curve fitting to a negative offset and it may affect the
appearance
on the monitor.
Please search in this forum for related topics. We ha longer
discussions
about.

Best regards --Gernot Hoffmann

Back to top

Ads

Advertising
Sponsor

Rodney
Guest

Posted: Mon Jun 16, 2008 10:41 am   Post subject: Re: Accurate formula to measure display GAMMA !

On Jun 16, 11:48 am, Gernot Hoffmann <hoffm...@fho-emden.de> wrote:

Quote:

Rodney schrieb:

On Jun 13, 7:41ï¿½pm, Gernot Hoffmann <hoffm...@fho-emden.de> wrote:
Rodney schrieb:

Dear Gernot,
Please find below the updated table, Please share your views for the
same
//////////
Input Stimulus
0
0.12549
0.25098
0.376471
0.501961
0.627451
0.752941
0.878431
1
//////////
Output Stimulus
0.00092
0.002908
0.019715
0.089258
0.203187
0.353551
0.532005
0.744232
1
/////////

Also what approach are you following ?
Rodney

Rodney,

it's the mathematical model Type B here on p.3:http://www.fho-emden.de/~hoffmann/measgamma10022004.pdf
The curve fitting is based on numerical optimization (Steepest
Descent), as explained.

Your actual data deliver Gamma=2.175 and Offset=-0.023.
This says: your monitor is too dark at the dark end. Dark
input values will be clipped visually.

Assumed, your white has 100cd/m2, then your black would be
about 0.1cd/m2. More realistic is 0.3cd/m2.

Feel free to send new measured data after a re-calibration
concerning the blackpoint.

Best regards --Gernot Hoffmann- Hide quoted text -

- Show quoted text -

Dear Gernot,
Below is the monitor current measurement
Brightness: 213 cd/m2
Black level: 0.2 cd/m2
CT: 6500
///////
Input
0.000000
0.125000
0.250000
0.375000
0.500000
0.625000
0.750000
0.875000
0.996094
////
Output
0.000857
0.004562
0.026489
0.088656
0.183950
0.326179
0.518590
0.744004
1.000000
////
B/w I hope you are using "single dimension steepest descent" algo as u
mentioned in the previous mail. Can you share xls sheet implementing
the same? Or C/C++ implementation?
Thanks,
Rodney

Rodney,

I'm using the algorithm chapter 4.2,Steepest descent with one-dimen-
sional search.
C/C++ is not a available, only the algorithm and PostScript code.
If you have Photoshop or InDesign or Illustrator or Ghostscript, then
you can use my Postscript code Reference [5] (rename as EPS).
Just replace the one table GamWht for White (gray) by your data, using
e.g. WordPad (plain text). Input left, output right.

Here is the result for your recent data:http://www.fho-emden.de/~hoffmann/rodney16062008.gif

Note that my program ignores the first value (at 0.0), because this is
mostly uncertain. The next value is IMHO still too dark which leads
in
the curve fitting to a negative offset and it may affect the
appearance
on the monitor.
Please search in this forum for related topics. We ha longer
discussions
about.

Best regards --Gernot Hoffmann

Thanks Gernot, That was quite helpful.
Rodney

Back to top

Ads

Advertising
Sponsor

Gernot Hoffmann
Guest

Posted: Mon Jun 16, 2008 2:41 pm   Post subject: Re: Accurate formula to measure display GAMMA !

Rodney schrieb:

Quote:

On Jun 16, 11:48 am, Gernot Hoffmann <hoffm...@fho-emden.de> wrote:
Rodney schrieb:

On Jun 13, 7:41ï¿½pm, Gernot Hoffmann <hoffm...@fho-emden.de> wrote:
Rodney schrieb:

Dear Gernot,
Please find below the updated table, Please share your views for the
same
//////////
Input Stimulus
0
0.12549
0.25098
0.376471
0.501961
0.627451
0.752941
0.878431
1
//////////
Output Stimulus
0.00092
0.002908
0.019715
0.089258
0.203187
0.353551
0.532005
0.744232
1
/////////

Also what approach are you following ?
Rodney

Rodney,

it's the mathematical model Type B here on p.3:http://www.fho-emden.de/~hoffmann/measgamma10022004.pdf
The curve fitting is based on numerical optimization (Steepest
Descent), as explained.

Your actual data deliver Gamma=2.175 and Offset=-0.023.
This says: your monitor is too dark at the dark end. Dark
input values will be clipped visually.

Assumed, your white has 100cd/m2, then your black would be
about 0.1cd/m2. More realistic is 0.3cd/m2.

Feel free to send new measured data after a re-calibration
concerning the blackpoint.

Best regards --Gernot Hoffmann- Hide quoted text -

- Show quoted text -

Dear Gernot,
Below is the monitor current measurement
Brightness: 213 cd/m2
Black level: 0.2 cd/m2
CT: 6500
///////
Input
0.000000
0.125000
0.250000
0.375000
0.500000
0.625000
0.750000
0.875000
0.996094
////
Output
0.000857
0.004562
0.026489
0.088656
0.183950
0.326179
0.518590
0.744004
1.000000
////
B/w I hope you are using "single dimension steepest descent" algo as u
mentioned in the previous mail. Can you share xls sheet implementing
the same? Or C/C++ implementation?
Thanks,
Rodney

Rodney,

I'm using the algorithm chapter 4.2,Steepest descent with one-dimen-
sional search.
C/C++ is not a available, only the algorithm and PostScript code.
If you have Photoshop or InDesign or Illustrator or Ghostscript, then
you can use my Postscript code Reference [5] (rename as EPS).
Just replace the one table GamWht for White (gray) by your data, using
e.g. WordPad (plain text). Input left, output right.

Here is the result for your recent data:http://www.fho-emden.de/~hoffmann/rodney16062008.gif

Note that my program ignores the first value (at 0.0), because this is
mostly uncertain. The next value is IMHO still too dark which leads
in
the curve fitting to a negative offset and it may affect the
appearance
on the monitor.
Please search in this forum for related topics. We ha longer
discussions
about.

Best regards --Gernot Hoffmann

Thanks Gernot, That was quite helpful.
Rodney

Rodney,

you're welcome. Checking your input data I found a
tiny mistake in my concept.
The normalized input data are 0, 1/8, 2/8, ...,8/8.
This is OK in PostScript, but finally we need integer
numbers. For instance: 4/8 delivers 255/2=127.5 as
input value in the table, but 128 for the color value of
the test pattern.
I would now choose
xi = 0 32 64 96 128 160 192 224 255
The normalized values are then Xi = xi / 255.
The last value is an exception.
These numbers are used for the generation of the
test patterns and for the curve fitting. Therefore I cannot
repair the bug immediately.
Considering all the uncertainties, the mistake isn't very
important, at the moment.

Best regards --Gernot Hoffmann

Back to top

Ads

Advertising
Sponsor

Gernot Hoffmann
Guest

Posted: Thu Jun 19, 2008 5:33 pm   Post subject: Re: Accurate formula to measure display GAMMA !

Gernot Hoffmann schrieb:

Quote:

Rodney schrieb:
On Jun 16, 11:48 am, Gernot Hoffmann <hoffm...@fho-emden.de> wrote:
Rodney schrieb:

On Jun 13, 7:41ï¿½pm, Gernot Hoffmann <hoffm...@fho-emden.de> wrote:
Rodney schrieb:

Dear Gernot,
Please find below the updated table, Please share your views for the
same
//////////
Input Stimulus
0
0.12549
0.25098
0.376471
0.501961
0.627451
0.752941
0.878431
1
//////////
Output Stimulus
0.00092
0.002908
0.019715
0.089258
0.203187
0.353551
0.532005
0.744232
1
/////////

Also what approach are you following ?
Rodney

Rodney,

it's the mathematical model Type B here on p.3:http://www.fho-emden.de/~hoffmann/measgamma10022004.pdf
The curve fitting is based on numerical optimization (Steepest
Descent), as explained.

Your actual data deliver Gamma=2.175 and Offset=-0.023.
This says: your monitor is too dark at the dark end. Dark
input values will be clipped visually.

Assumed, your white has 100cd/m2, then your black would be
about 0.1cd/m2. More realistic is 0.3cd/m2.

Feel free to send new measured data after a re-calibration
concerning the blackpoint.

Best regards --Gernot Hoffmann- Hide quoted text -

- Show quoted text -

Dear Gernot,
Below is the monitor current measurement
Brightness: 213 cd/m2
Black level: 0.2 cd/m2
CT: 6500
///////
Input
0.000000
0.125000
0.250000
0.375000
0.500000
0.625000
0.750000
0.875000
0.996094
////
Output
0.000857
0.004562
0.026489
0.088656
0.183950
0.326179
0.518590
0.744004
1.000000
////
B/w I hope you are using "single dimension steepest descent" algo as u
mentioned in the previous mail. Can you share xls sheet implementing
the same? Or C/C++ implementation?
Thanks,
Rodney

Rodney,

I'm using the algorithm chapter 4.2,Steepest descent with one-dimen-
sional search.
C/C++ is not a available, only the algorithm and PostScript code.
If you have Photoshop or InDesign or Illustrator or Ghostscript, then
you can use my Postscript code Reference [5] (rename as EPS).
Just replace the one table GamWht for White (gray) by your data, using
e.g. WordPad (plain text). Input left, output right.

Here is the result for your recent data:http://www.fho-emden.de/~hoffmann/rodney16062008.gif

Note that my program ignores the first value (at 0.0), because this is
mostly uncertain. The next value is IMHO still too dark which leads
in
the curve fitting to a negative offset and it may affect the
appearance
on the monitor.
Please search in this forum for related topics. We ha longer
discussions
about.

Best regards --Gernot Hoffmann

Thanks Gernot, That was quite helpful.
Rodney

Rodney,

you're welcome. Checking your input data I found a
tiny mistake in my concept.
The normalized input data are 0, 1/8, 2/8, ...,8/8.
This is OK in PostScript, but finally we need integer
numbers. For instance: 4/8 delivers 255/2=127.5 as
input value in the table, but 128 for the color value of
the test pattern.
I would now choose
xi = 0 32 64 96 128 160 192 224 255
The normalized values are then Xi = xi / 255.
The last value is an exception.
These numbers are used for the generation of the
test patterns and for the curve fitting. Therefore I cannot
repair the bug immediately.
Considering all the uncertainties, the mistake isn't very
important, at the moment.

Best regards --Gernot Hoffmann

The PDF and the program file were updated. A new realistic
example for Eizo CG19 (TFT monitor) was added.
http://www.fho-emden.de/~hoffmann/measgamma10022004.pdf

G.H.

Back to top

Ads

Advertising
Sponsor

Rodney
Guest

Posted: Sat Jun 21, 2008 5:24 am   Post subject: Re: Accurate formula to measure display GAMMA !

On Jun 19, 10:33 pm, Gernot Hoffmann <hoffm...@fho-emden.de> wrote:

Quote:

Gernot Hoffmann schrieb:

Rodney schrieb:
On Jun 16, 11:48 am, Gernot Hoffmann <hoffm...@fho-emden.de> wrote:
Rodney schrieb:

On Jun 13, 7:41ï¿½pm, Gernot Hoffmann <hoffm...@fho-emden.de> wrote:
Rodney schrieb:

Dear Gernot,
Please find below the updated table, Please share your views for the
same
//////////
Input Stimulus
0
0.12549
0.25098
0.376471
0.501961
0.627451
0.752941
0.878431
1
//////////
Output Stimulus
0.00092
0.002908
0.019715
0.089258
0.203187
0.353551
0.532005
0.744232
1
/////////

Also what approach are you following ?
Rodney

Rodney,

it's the mathematical model Type B here on p.3:http://www.fho-emden.de/~hoffmann/measgamma10022004.pdf
The curve fitting is based on numerical optimization (Steepest
Descent), as explained.

Your actual data deliver Gamma=2.175 and Offset=-0.023.
This says: your monitor is too dark at the dark end. Dark
input values will be clipped visually.

Assumed, your white has 100cd/m2, then your black would be
about 0.1cd/m2. More realistic is 0.3cd/m2.

Feel free to send new measured data after a re-calibration
concerning the blackpoint.

Best regards --Gernot Hoffmann- Hide quoted text -

- Show quoted text -

Dear Gernot,
Below is the monitor current measurement
Brightness: 213 cd/m2
Black level: 0.2 cd/m2
CT: 6500
///////
Input
0.000000
0.125000
0.250000
0.375000
0.500000
0.625000
0.750000
0.875000
0.996094
////
Output
0.000857
0.004562
0.026489
0.088656
0.183950
0.326179
0.518590
0.744004
1.000000
////
B/w I hope you are using "single dimension steepest descent" algo as u
mentioned in the previous mail. Can you share xls sheet implementing
the same? Or C/C++ implementation?
Thanks,
Rodney

Rodney,

I'm using the algorithm chapter 4.2,Steepest descent with one-dimen-
sional search.
C/C++ is not a available, only the algorithm and PostScript code.
If you have Photoshop or InDesign or Illustrator or Ghostscript, then
you can use my Postscript code Reference [5] (rename as EPS).
Just replace the one table GamWht for White (gray) by your data, using
e.g. WordPad (plain text). Input left, output right.

Here is the result for your recent data:http://www.fho-emden.de/~hoffmann/rodney16062008.gif

Note that my program ignores the first value (at 0.0), because this is
mostly uncertain. The next value is IMHO still too dark which leads
in
the curve fitting to a negative offset and it may affect the
appearance
on the monitor.
Please search in this forum for related topics. We ha longer
discussions
about.

Best regards --Gernot Hoffmann

Thanks Gernot, That was quite helpful.
Rodney

Rodney,

you're welcome. Checking your input data I found a
tiny mistake in my concept.
The normalized input data are 0, 1/8, 2/8, ...,8/8.
This is OK in PostScript, but finally we need integer
numbers. For instance: 4/8 delivers 255/2=127.5 as
input value in the table, but 128 for the color value of
the test pattern.
I would now choose
xi = 0 32 64 96 128 160 192 224 255
The normalized values are then Xi = xi / 255.
The last value is an exception.
These numbers are used for the generation of the
test patterns and for the curve fitting. Therefore I cannot
repair the bug immediately.
Considering all the uncertainties, the mistake isn't very
important, at the moment.

Best regards --Gernot Hoffmann

The PDF and the program file were updated. A new realistic
example for Eizo CG19 (TFT monitor) was added.http://www.fho-emden.de/~hoffmann/measgamma10022004.pdf

G.H.

Dear Gernot,

I was going through your algorithm mentioned in
http://www.fho-emden.de/~hoffmann/measgamma10022004.pdf
4.2 Mathematics ( One dimensional search by comparison of ten samples)

and found a small issue in it.

Under "One dimensional search"
q1 = p1
q2 = p2
////
Fm = 999 ----------> ISSUE
///

I suppose it should be "Fm = Fo" instead of "Fm = 999"

PLEASE clarify.
Thank-you,
Rodney

Back to top

Ads

Advertising
Sponsor

Gernot Hoffmann
Guest

Posted: Sat Jun 21, 2008 6:09 am   Post subject: Re: Accurate formula to measure display GAMMA !

Rodney schrieb:

Quote:

On Jun 19, 10:33 pm, Gernot Hoffmann <hoffm...@fho-emden.de> wrote:
Gernot Hoffmann schrieb:

Rodney schrieb:
On Jun 16, 11:48 am, Gernot Hoffmann <hoffm...@fho-emden.de> wrote:
Rodney schrieb:

On Jun 13, 7:41ï¿½pm, Gernot Hoffmann <hoffm...@fho-emden..de> wrote:
Rodney schrieb:

Dear Gernot,
Please find below the updated table, Please share your views for the
same
//////////
Input Stimulus
0
0.12549
0.25098
0.376471
0.501961
0.627451
0.752941
0.878431
1
//////////
Output Stimulus
0.00092
0.002908
0.019715
0.089258
0.203187
0.353551
0.532005
0.744232
1
/////////

Also what approach are you following ?
Rodney

Rodney,

it's the mathematical model Type B here on p.3:http://www.fho-emden.de/~hoffmann/measgamma10022004.pdf
The curve fitting is based on numerical optimization (Steepest
Descent), as explained.

Your actual data deliver Gamma=2.175 and Offset=-0.023.
This says: your monitor is too dark at the dark end. Dark
input values will be clipped visually.

Assumed, your white has 100cd/m2, then your black would be
about 0.1cd/m2. More realistic is 0.3cd/m2.

Feel free to send new measured data after a re-calibration
concerning the blackpoint.

Best regards --Gernot Hoffmann- Hide quoted text -

- Show quoted text -

Dear Gernot,
Below is the monitor current measurement
Brightness: 213 cd/m2
Black level: 0.2 cd/m2
CT: 6500
///////
Input
0.000000
0.125000
0.250000
0.375000
0.500000
0.625000
0.750000
0.875000
0.996094
////
Output
0.000857
0.004562
0.026489
0.088656
0.183950
0.326179
0.518590
0.744004
1.000000
////
B/w I hope you are using "single dimension steepest descent" algo as u
mentioned in the previous mail. Can you share xls sheet implementing
the same? Or C/C++ implementation?
Thanks,
Rodney

Rodney,

I'm using the algorithm chapter 4.2,Steepest descent with one-dimen-
sional search.
C/C++ is not a available, only the algorithm and PostScript code.
If you have Photoshop or InDesign or Illustrator or Ghostscript, then
you can use my Postscript code Reference [5] (rename as EPS).
Just replace the one table GamWht for White (gray) by your data, using
e.g. WordPad (plain text). Input left, output right.

Here is the result for your recent data:http://www.fho-emden.de/~hoffmann/rodney16062008.gif

Note that my program ignores the first value (at 0.0), because this is
mostly uncertain. The next value is IMHO still too dark which leads
in
the curve fitting to a negative offset and it may affect the
appearance
on the monitor.
Please search in this forum for related topics. We ha longer
discussions
about.

Best regards --Gernot Hoffmann

Thanks Gernot, That was quite helpful.
Rodney

Rodney,

you're welcome. Checking your input data I found a
tiny mistake in my concept.
The normalized input data are 0, 1/8, 2/8, ...,8/8.
This is OK in PostScript, but finally we need integer
numbers. For instance: 4/8 delivers 255/2=127.5 as
input value in the table, but 128 for the color value of
the test pattern.
I would now choose
xi = 0 32 64 96 128 160 192 224 255
The normalized values are then Xi = xi / 255.
The last value is an exception.
These numbers are used for the generation of the
test patterns and for the curve fitting. Therefore I cannot
repair the bug immediately.
Considering all the uncertainties, the mistake isn't very
important, at the moment.

Best regards --Gernot Hoffmann

The PDF and the program file were updated. A new realistic
example for Eizo CG19 (TFT monitor) was added.http://www.fho-emden.de/~hoffmann/measgamma10022004.pdf

G.H.

Dear Gernot,

I was going through your algorithm mentioned in
http://www.fho-emden.de/~hoffmann/measgamma10022004.pdf
4.2 Mathematics ( One dimensional search by comparison of ten samples)

and found a small issue in it.

Under "One dimensional search"
q1 = p1
q2 = p2
////
Fm = 999 ----------> ISSUE
///

I suppose it should be "Fm = Fo" instead of "Fm = 999"

PLEASE clarify.
Thank-you,
Rodney

Rodney,

thanks for the hint. There is indeed a discrepancy between the
(really working) PostScript code with Fm=Fo and the algorithmical
explanation with Fm=999. The PostScript code is valid, but the
other version would IMO work as well.
Explanation:
The number 999 indicates a nonsensical large starting value
for the function F which has to be minimized. Any executed step
will deliver a smaller value. Fo and Fu are the previous and the
actual
values for a global steepest descent step.
Fm is the starting value for the one-dimensional search. The
direction is fixed (no new gradient) but the stepsize has to be
adjusted. This is done by an extremely robust method: subdivide
the global stepsize (which is often too large) into 10 steps and
find the minimum along the trajectory. Under normal circumstances
this would work as well with Fm=999.
Please follow in any case of doubt the PostScript version (new since
a few days, same as in the updated doc).

Best regards --Gernot Hoffmann

Back to top

Ads

Advertising
Sponsor

Rodney
Guest

Posted: Sat Jun 21, 2008 8:13 am   Post subject: Re: Accurate formula to measure display GAMMA !

On Jun 21, 11:09 am, Gernot Hoffmann <hoffm...@fho-emden.de> wrote:

Quote:

Rodney schrieb:

On Jun 19, 10:33 pm, Gernot Hoffmann <hoffm...@fho-emden.de> wrote:
Gernot Hoffmann schrieb:

Rodney schrieb:
On Jun 16, 11:48 am, Gernot Hoffmann <hoffm...@fho-emden.de> wrote:
Rodney schrieb:

On Jun 13, 7:41ï¿½pm, Gernot Hoffmann <hoffm...@fho-emden.de> wrote:
Rodney schrieb:

Dear Gernot,
Please find below the updated table, Please share your views for the
same
//////////
Input Stimulus
0
0.12549
0.25098
0.376471
0.501961
0.627451
0.752941
0.878431
1
//////////
Output Stimulus
0.00092
0.002908
0.019715
0.089258
0.203187
0.353551
0.532005
0.744232
1
/////////

Also what approach are you following ?
Rodney

Rodney,

it's the mathematical model Type B here on p.3:http://www.fho-emden.de/~hoffmann/measgamma10022004.pdf
The curve fitting is based on numerical optimization (Steepest
Descent), as explained.

Your actual data deliver Gamma=2.175 and Offset=-0.023.
This says: your monitor is too dark at the dark end. Dark
input values will be clipped visually.

Assumed, your white has 100cd/m2, then your black would be
about 0.1cd/m2. More realistic is 0.3cd/m2.

Feel free to send new measured data after a re-calibration
concerning the blackpoint.

Best regards --Gernot Hoffmann- Hide quoted text -

- Show quoted text -

Dear Gernot,
Below is the monitor current measurement
Brightness: 213 cd/m2
Black level: 0.2 cd/m2
CT: 6500
///////
Input
0.000000
0.125000
0.250000
0.375000
0.500000
0.625000
0.750000
0.875000
0.996094
////
Output
0.000857
0.004562
0.026489
0.088656
0.183950
0.326179
0.518590
0.744004
1.000000
////
B/w I hope you are using "single dimension steepest descent" algo as u
mentioned in the previous mail. Can you share xls sheet implementing
the same? Or C/C++ implementation?
Thanks,
Rodney

Rodney,

I'm using the algorithm chapter 4.2,Steepest descent with one-dimen-
sional search.
C/C++ is not a available, only the algorithm and PostScript code.
If you have Photoshop or InDesign or Illustrator or Ghostscript, then
you can use my Postscript code Reference [5] (rename as EPS).
Just replace the one table GamWht for White (gray) by your data, using
e.g. WordPad (plain text). Input left, output right.

Here is the result for your recent data:http://www.fho-emden.de/~hoffmann/rodney16062008.gif

Note that my program ignores the first value (at 0.0), because this is
mostly uncertain. The next value is IMHO still too dark which leads
in
the curve fitting to a negative offset and it may affect the
appearance
on the monitor.
Please search in this forum for related topics. We ha longer
discussions
about.

Best regards --Gernot Hoffmann

Thanks Gernot, That was quite helpful.
Rodney

Rodney,

you're welcome. Checking your input data I found a
tiny mistake in my concept.
The normalized input data are 0, 1/8, 2/8, ...,8/8.
This is OK in PostScript, but finally we need integer
numbers. For instance: 4/8 delivers 255/2=127.5 as
input value in the table, but 128 for the color value of
the test pattern.
I would now choose
xi = 0 32 64 96 128 160 192 224 255
The normalized values are then Xi = xi / 255.
The last value is an exception.
These numbers are used for the generation of the
test patterns and for the curve fitting. Therefore I cannot
repair the bug immediately.
Considering all the uncertainties, the mistake isn't very
important, at the moment.

Best regards --Gernot Hoffmann

The PDF and the program file were updated. A new realistic
example for Eizo CG19 (TFT monitor) was added.http://www.fho-emden.de/~hoffmann/measgamma10022004.pdf

G.H.

Dear Gernot,

I was going through your algorithm mentioned in
http://www.fho-emden.de/~hoffmann/measgamma10022004.pdf
4.2 Mathematics ( One dimensional search by comparison of ten samples)

and found a small issue in it.

Under "One dimensional search"
q1 = p1
q2 = p2
////
Fm = 999 ----------> ISSUE
///

I suppose it should be "Fm = Fo" instead of "Fm = 999"

PLEASE clarify.
Thank-you,
Rodney

Rodney,

thanks for the hint. There is indeed a discrepancy between the
(really working) PostScript code with Fm=Fo and the algorithmical
explanation with Fm=999. The PostScript code is valid, but the
other version would IMO work as well.
Explanation:
The number 999 indicates a nonsensical large starting value
for the function F which has to be minimized. Any executed step
will deliver a smaller value. Fo and Fu are the previous and the
actual
values for a global steepest descent step.
Fm is the starting value for the one-dimensional search. The
direction is fixed (no new gradient) but the stepsize has to be
adjusted. This is done by an extremely robust method: subdivide
the global stepsize (which is often too large) into 10 steps and
find the minimum along the trajectory. Under normal circumstances
this would work as well with Fm=999.
Please follow in any case of doubt the PostScript version (new since
a few days, same as in the updated doc).

Best regards --Gernot Hoffmann

Dear Gernot,

Yes, I was in process of implementing the same in C, and this found
this small issue. IF we use Fm=999, the gamma value would not be
accurate. After changing Fm = Fo, the results were quite good.

I have a small query:
1. Why only taking 9 steps for input stimulus 0, .125, ..1.0 ???? Why
not all 255 steps?
2. Why dividing the global stepsize into 10 only??

Thank-you,
Rodney

Back to top

Ads

Advertising
Sponsor

Display posts from previous:

       Physics Talk Forum Index -> Color and Light All times are GMT
Goto page 1, 2  Next

Page 1 of 2

Jump to:

You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum

Australian Debt Consolidation Experts
medical insurance
Wedding Website
Search Escorts and girls (incall/OutCall) online with Oasi2000, ItaliaChiamami, BestAnnunci...
Sex Pals
Xpointer Framework
cheap life insurance
Make Your Own Website
Cheap phone calls to UAE Long island Cleaning service
toxic mold
UK Swingers Genuine Contacts Site
cleaning supplies
bissell Parts