How can i count the number of pixels for a specific color using the L*A*B?

I want to write a code that will count the number of pixels for all the tints of a specific color in an image.

Answer 1

[Image Analyst]

Just make a mask and use nnz():

lMask = lImage > Lmin & lImage <= Lmax;
aMask = aImage > Amin & aImage <= Amax;
bMask = bImage > Bmin & bImage <= Bmax;
overallMask = lMask & aMask & bMask;
numberOfPixels = nnz(overallMask)

Of you can make a 3-D histogram.

Look at colorcloud().

[Stephani Kanga]

can you show me an example become i'm new here and i don't know much about matlab?

[Image Analyst]

With what image? You forgot to attach it.

Steve shows one way in his Blog:

https://www.mathworks.com/help/images/color-based-segmentation-using-k-means-clustering.html?prodcode=IP&language=en

Try to adapt that.

[Stephani Kanga]

Yes for example i have this image and i want to calculate the % of the pixels that have dark purple-black color.

[Image Analyst]

OK. So did you do it? The demo gives you the class for each pixel so just sum them up and divide by the number of pixels and multiply by 100 to get percent area for each class.

classCounts = histcounts(pixel_labels)
percentAreas = 100 * classCounts / numel(pixel_labels)

The tricky part is you have to know which class is dark purple but you can look at the cluster centers to get the color of each class and then you can figure out what class number dark purple is. That class number could possibly vary each time you run the program because kmeans uses random numbers.