# Skin color detection problem

61 views (last 30 days)
matlab_image on 9 Jul 2012
Commented: Image Analyst on 27 Sep 2018
I have following code of skin detection but I am not understanding it completely. Is it HSV or Ycbcr or combined. Can anybody please tell me which type of this code is. And please explain me following code in brief.
I=double(I);
[hue,s,v]=rgb2hsv(I);
cb = 0.148* I(:,:,1) - 0.291* I(:,:,2) + 0.439 * I(:,:,3) + 128;
cr = 0.439 * I(:,:,1) - 0.368 * I(:,:,2) -0.071 * I(:,:,3) + 128;
[w h]=size(I(:,:,1));
for i=1:w
for j=1:h
if 140<=cr(i,j) && cr(i,j)<=165 && 140<=cb(i,j) && cb(i,j)<=195 && 0.01<=hue(i,j) && hue(i,j)<=0.1
segment(i,j)=1;
else
segment(i,j)=0;
end
end
end
% imshow(segment);
im(:,:,1)=I(:,:,1).*segment;
im(:,:,2)=I(:,:,2).*segment;
im(:,:,3)=I(:,:,3).*segment;
figure,imshow(uint8(im));

zonaira afzal on 16 Nov 2017
I don't want to black the other area i want it to be same as in the original picture. what i should do? i want to dress up the a person, and don't want the dress to be apply on the hands feet, neck and head.. Kindly help me in this regard...
Ravindra Singh Pawar on 25 Sep 2018
Image Analyst on 25 Sep 2018
I seriously doubt whether that ad hoc code was the subject of any research paper. For papers on image processing, see http://www.visionbib.com/bibliography/contents.html - it has almost all of them.

Image Analyst on 9 Jul 2012
It's using both - for some reason. It's using Cb, Cr, and hue.

Ravindra Singh Pawar on 27 Sep 2018
how can we not write ycbcr like [y,cb,cr]=rgb2ycbcr(I);
Image Analyst on 27 Sep 2018
Not sure, that's just the way it is. Actually using rgb2hsv() like
[hue,s,v]=rgb2hsv(rgbImage);
is an undocumented usage. The help for rgb2hsv() does not mention the possibility of accepting 3 returned arrays. It only mentions that one array is returned.

Karbala'a Unvi. Science on 14 Oct 2013
Dear Sir, I hope that you help me in skin detection by using the RGB image directly not transfer it in to the Cb Cr color model

#### 1 Comment

Image Analyst on 14 Oct 2013
Skin color cannot be segmented in RGB color space by simple thresholding. That's because the gamut in 3D color space is not a simple box shape that can be carved out by thresholding. At least not for skin in general, though you may be able to do it for one person's skin, but not for skin of multiple different people of different skin colors. If you were to do it in RGB space you'd have to employ a complicated ad hoc scheme like tree bagging or something. That's much more complicated than going to a different color space though it would probably do a better job than simple thresholding in YCbCr space because it can follow the curves of the boomerang-shaped gamut better. Of course using tree bagger would work in any space due to the nature of it, but it depends on a complete training set where you teach it with all skin colors you ever expect to encounter. I know it's very confusing but I hope that explains it better.