So...you want to remove the fat white borders from around the black region??
You could try search for the first and last row,column that are 0 in the second image, using (e.g.)
% assuming BW2 is 2D, one channel.
% the reason the first has a BW2' and the second is BW
% is because of Matlab being column-major when it does `find`:
% have to make sure we actually get the first/last indices.
[col1,row1] = find(BW2'==0,1,'first'); % top-left corner
[row2,col2] = find(BW2==0,1,'last'); % bottom-right corner
BW3 = BW(row1:row2, col1:col2);
This does assume that the top-left and bottom-right corner of the image proper won't be part of the outline - this is a very low chance.
If you want to be sure however, you could instead look for the first row that is not all 255 and first col that is not all 255, and similarly for the last row/col (ie locate margin by looking for entire rows/cols being 255) and crop like that:
cols = sum(abs(255-BW2),1); % work out difference from 255
rows = sum(abs(255-BW2),2);
col1 = find(cols,1,'first');
col2 = find(cols,1,'last');
row1 = find(rows,1,'first');
row2 = find(rows,1,'last');
BW3 = BW(row1:row2,col1:col2);
(Note this will fail on your jpeg because of the compression it uses -- the border is not uniquely 255, there are a few 251's in there. But it will work on your code if you don't save
BW out as a jpeg first and then read it back in. If you're going to save images, use a lossless format like TIF or PNG).