# detect square in image

173 views (last 30 days)
Abdul on 29 Dec 2011
Commented: Dmitry Surov on 16 Sep 2020
i want to detect square shape only in image. how can i do this using matlab?

Chandra Kurniawan on 29 Dec 2011
Hello,
Is your goal only to detect square in this picture?
If so, you don't need to do shape recognition, etc.
You just need to perform some morphologichal operations only.
Ibw = ~im2bw(I,graythresh(I));
Ifill = imfill(Ibw,'holes');
Iarea = bwareaopen(Ifill,100);
Ifinal = bwlabel(Iarea);
stat = regionprops(Ifinal,'boundingbox');
imshow(I); hold on;
for cnt = 1 : numel(stat)
bb = stat(cnt).BoundingBox;
rectangle('position',bb,'edgecolor','r','linewidth',2);
end
And the result :
Is this what you need?
Dmitry Surov on 16 Sep 2020
Can you please explain what do we get in stat variable ? It's something related to border line of each square, but i misunderstand what that means
Thanks

### More Answers (2)

Sean de Wolski on 29 Dec 2011
Those squares are very well defined.
• Set a less than threshold to make the dark colors true and everything else false. ( BW = I<thresh )
• Perform a connected components analysis ( doc bwconncomp)
• Then regionprops with EulerNumber specified.
• Apply a threshold to the regions matching the EulerNumber ( idx = [stats(:).EulerNumber] == x ) of a sample square.
• Build a new matrix of false. ( BWnew = false(CC.ImageSize))
• Set the indices in the connected components analysis corresponding to the matching EulerNumber to true. ( BWnew(vertcat(CC.PixelIdxList{idx})) = true)
In theory you'll have a mask of the boxes.

Naz on 29 Dec 2011
Edited: John Kelly on 14 Nov 2013
I just thought of an alternative idea. Instead of detecting a square, you should register the image (filled form) of interest with the raw image. Then, since you know the coordinates of each square, you can just check the pixels at those coordinates for the white/black values.