crop.m

function [ mask ] = crop( eyeregion )
%CROP Crop eye region based on red channel thresholding with largest
%connected component
% 
% RETURN
%   A mask that is 1 if it is in the eye region

threshold = 57;
r = eyeregion(:, :, 1);
mask = r > threshold;

% the second largest connected component is the eye region
conncomp = bwconncomp(mask);
numPixels = cellfun(@numel, conncomp.PixelIdxList);
[~, i1] = max(numPixels);
numPixels(i1) = 0;
[~, i2] = max(numPixels);
mask = zeros(size(r));
mask(conncomp.PixelIdxList{i2}) = 1;

% eye region has to be one piece without holes

% eliminate false negative
conncomp = bwconncomp(1 - mask);
numPixels = cellfun(@numel, conncomp.PixelIdxList);

[~, i1] = max(numPixels);
inds = 1: length(numPixels);
inds(i1) = [];
for i = 1: length(inds)
    mask(conncomp.PixelIdxList{inds(i)}) = 1;
end

% % eliminate false positive
% inds = 1: length(numPixels);
% inds(i1) = [];
% for i = 1: length(inds)
%     mask(conncomp.PixelIdxList{inds(i)}) = 0;
% end
% 


end