Skin Pixel Likelihoods and Skin Detection
-----------------------------------------------

This Matlab code was developed for skin pixel detection in general imagery.

Non-parametric histogram-based models were trained using manually annotated
skin and non-skin pixels. A total of 14,985,845 skin pixels and 304,844,751
non-skin pixels were used.


I hope you find it useful.
				- Ciarán Ó Conaire (oconaire at eeng.dcu.ie)


Demo:

- Start Matlab
- change to the folder that contains TestModel.m
- run TestModel.m

The demo requires the following files to be in the same folder:
	computeSkinProbability.m
	skinmodel.bin
	normalise.m


General Usage:

	% Load an RGB image
	im = imread('image001.jpg');

	% convert the image data to doubles
	im = double(im);

	% compute the skin likelihood for each pixel
	skinprob = computeSkinProbability(im);

	% threshold the likelihood to detect skin
	skin = (skinprob > 0)+0; % zero is added to cast from a logical-typed to a double-typed matrix


	Note that the following files should be in the path:
		computeSkinProbability.m
		skinmodel.bin



---------------------------------------
More info:


The detector works as follows:
- I gather a collection of images and mark each pixel as "skin" or "non skin"
- I create an RGB histogram for "skin pixels" and another one for "non-skin pixels"
(they are 32x32x32 in size)
- For a particular bin (i.e. pixel colour), the log likelihood of it being skin is

log(H(r,g,b) / h(r,g,b))

...where H is the skin histogram, and h is the non-skin histogram

For a new image, I compute the log likelihood of each pixel and then threshold the result to decide skin/non-skin.