ColorChecker function does not work?

I am trying to follow the example here:

Comparison of Auto White Balance Algorithms - MATLAB & Simulink (mathworks.com)

I use this code:

cfa = rawread("DSC_0136.NEF"); 
J = demosaic(cfa,"rggb");
A_sRGB = lin2rgb(J);
montage({J,A_sRGB})
title("Original Image Before and After Gamma Correction")
chart_sRGB = colorChecker(A_sRGB);
% chart_sRGB = colorChecker(rgb, sensitivity=0);
DisplayChart(chart_sRGB)

But I keep running into this error:

Error using colorChecker

Color patches were not detected. Try adjusting the Sensitivity parameter or turning off the Downsample parameter.

This is my test NEF:

Why is the functioon failing?

I tried varying the sensitivity from 0 to 1 and turning downsampling off to no avail.

Is the function not designed to operate on a picture of the chart that's full-frame?

[Roger Breton]

BTW, it was painful to have to post the question twice because the "Captcha" did not work?

[Walter Roberson]

Please zip the .NEF file and attach the .NEF.zip

[Roger Breton]

Out of curiosity, to test the hypothesis that the ColorChecker function needs a smaller size chart in the image, I took this picture this morning:

As you can see, the chart occupies a small area compared to the size of the image. YET, when running my code, I still the same error message:

Error using colorChecker: Color patches were not detected. Try adjusting the Sensitivity parameter or turning off the Downsample parameter.

[Roger Breton]

With this statement: chart_sRGB = colorChecker(scaled_img, sensitivity=0);

I get: Color patches were not detected. Try adjusting the Sensitivity parameter or turning off the Downsample parameter.

But with sensitivity=1, I this other error message:

Color chart was not detected in the image. Try adjusting input parameters or manually specifying the Registration Points.

Which begs the question, how can I possibly manuall specify the Registration points? I looked at the Registration Point option last night and didn't find anything I could do manually.

oh well, Rome wasn't built in one day...

[Roger Breton]

Conduction a search with "ColorChecker" turned a few interesting link, such as this one:

box detection/rectangle detection - MATLAB Answers - MATLAB Central (mathworks.com)

Naturally, I ran the code and it worked:

clc; clear all; close all;
filename = 'https://www.mathworks.com/matlabcentral/answers/uploaded_files/812319/image.jpeg';
img = imread(filename);
% use colorChecker
chart = colorChecker(img);
displayChart(chart)

-

I downloaded the chart and could not see how much "different" it was from the my two chart images?

[Roger Breton]

I tried all kinds of things and the only version of the ColorChecker chart that works is the one from MathWorks:

https://www.mathworks.com/matlabcentral/answers/uploaded_files/812319/image.jpeg

I tried adding white crosses to my charts to no avail?

I tried making the edges of my chart R=G=B=0 to no avail?

I give up.

[Roger Breton]

I tried to trace through execution in the ColorChecker.m but could not find traces of the particular error message I receive :(

[Roger Breton]

I removed the four little "gray" pieces from the chart and it still does not work.

[Roger Breton]

From the ColorChecker documentation, I discovered that it was possible to manually detect the chart by supplying the corner points? So I used the MathWorks ColorChecker chart that "works", giving this output:

chart =

colorChecker with properties:

Image: [876×1312×3 uint8]

RegistrationPoints: [4×2 double]

ColorROIs: [24×1 struct]

... to try to "reverse-engineer" the chart object?

I imported the image into Adobe Illustrator and looked up the "RegistrationPoints" array onto the image. Using these coordinates, I tried to create the chart object using this code:

cornerPoints = [1113,632;638,636;632,323;1110,320];

chart = colorChecker(rgbImage,"RegistrationPoints",cornerPoints);

This method may not be elegant but if it allows me to experiment with the more advanced functions I want to get at like "measure WhiteBalance", it's worth the effort! So here is my result:

I should not have to work this hard...

[Roger Breton]

The RegistrationPoints order is:

[Lower Right X, Lower Right Y; Lower Left X, Lower Left Y; Upper Left X, Upper Left Y; Upper Right X, Upper Right Y]