-
|
Hello, I'm looking for a help, My goal is to use the spectral reflectance of colorchecker in place of reference chart under controlled light for color corrections. Collect under noncontroled light, the reflectance of colorchecker, to use as reference in color correction. however, it donot work as expected, perhaps, i'm missing something. this image has a resume of the approach:
here is the code: Using COLOUR - 0.4.1the CCT in the environmet, was about 4200Kselected Illuminant FL2 -- since is the one closer to 4200K CCTConverted the DNG to tiif
Loop over each reflectance to build the reference XYZ, used the FL2 as illuminanthere, is possible to replace the CMFS by spectral sensitivity of camera?for example, iphone spectral sensitivity is available here: http://ohlab.kic.ac.jp/data/iPhoneX.txt**I tried also this approach, used the spectral sensitivity of iphone, and got worst results!`dicionario = [c1,c2,c3,c4,c5,c6,c7,c8,c9,c10,c11,c12,c13,c14,c15,c16,c17,c18,c19,c20,c21,c22,c23,c24] illumini = np.array([0.36888,0.37556]) for i in dicionario: this is the ouput, so adopted as REFERENCE:
After used the colour checker dectetion lib, identified the the swatches, and applied the rREFERENCE above to correct the image.`RGB = colour.read_image('/home/user1/phd/novafotos/t1/linearf.tiff') REFERENCE_SWATCHES ##reshape the SWATCHES ##CCM method to new Imagenewimg = colour.colour_correction(RGB, newarr, REFERENCE_SWATCHES); print ("Cor corrigida") plot real image.print ("Cor Original") the color correct output was:
DNG Image is available here:https://drive.google.com/drive/folders/1ebZYz8FceQDrMhEWAs1eYhB26dFpxUvk?usp=sharing reflectance data from colorcheck is available here:
|
Beta Was this translation helpful? Give feedback.
Replies: 2 comments 3 replies
-
|
Hi @xeggbee, When you generate the reference values and perform the Spectral integration with the Cheers, Thomas |
Beta Was this translation helpful? Give feedback.
-
thanks! thanks thanks! it really changed!
sounds like that the red is impacted.
i'll try to iterate over each illuminant and see if changes, i'm thinking like this because if I perform the color correction (ignoring the spectral data) using the standard D65, performs much better in the color correction.
as next step, i'll try the steps below:So, in poor understanding, the spectral correction might be impacted due the illuminant, due the DNG to tiff conversion (gamma and 8 or 16 bits output) or camera white balance (AWB, Tugsten, Fluorescent, Daylight and Cloudy). Please let know if have some suggestion over it. others concers: Where is the moment that should use the Spectral Sensitivity Function?Another point is, for this flow, where is the moment that i can use the Spectral Sensitivy function (SSF) of camera? is during the sd_to_XYZ in place of CMFS?
Scene and CC reflectance.Checked your comment in #747747#discussioncomment-287904_ and you suggest to also capture the spectral irradiance of the scene. **for this data collection used the spectral radiometer Field Spec 4. Also reported to use the spectro-photometer to capture spectral reflectance of the ColorChecker, my question here is, for case of non standard light source, does the spectro-photometer also works? (its much cheapper than spectralradiometer) spectral irradiance of the scene and the spectral reflectance of the ColorChecker Classic (with a spectro-photometer) |
Beta Was this translation helpful? Give feedback.
-
If you are using a chart based workflow, you don't really need to use camera sensitivities, those do not require a chart because you can generate reference values at will and at any desired white-balance. We have a lot of relevant code for that here: https://github.com/colour-science/colour/blob/develop/colour/characterisation/aces_it.py and it is being used there for example: https://github.com/ampas/idt-calculator/
It depends on your device but you would probably either want a cosine corrector to sample the hemisphere above the chart or alternatively, measure a reflectance standard such as Spectralon where your chart is located. You could then use it to generate references values to match for your chart. |
Beta Was this translation helpful? Give feedback.
-
|
thanks again! Tried the Aces, but not sure where to go now.
[ 2.15243805 1. 1.81062535] #plot got ping
how to move to other white-balance?how to export back to sRGB? |
Beta Was this translation helpful? Give feedback.
-
Am I placing the Spectral Referece at the right place? [using the example from colour-checker-detection]Delta E seems pretty good to be true! This is the SD to XYZ
In place of REFERENCE_SWATCHES, used the XYZ1 (the SD to XYZ)`for i, swatches in enumerate(SWATCHES): ` `###Before Correction ##dAfter COrrection
Before CCM - delta E After CCM - delta E Some suggestion to determine the SDS_ILLUMINANTS when capture was in non-controlled light?I'm trying two approaches: 2-delta E |
Beta Was this translation helpful? Give feedback.
Hi @xeggbee,
When you generate the reference values and perform the Spectral integration with the
colour.sd_to_XYZdefinition, you should try to divide by 100 because the values returned by default are expressed between 0 and 100 as percentage in the reflectance case.Cheers,
Thomas