Identification of rooms in a floorplan

I have a floorplan image with only walls. I want to identify the room coordinates of the floorplan.

This code actually closes the doors and identifies the rooms in the floorplan. This works well when shape of the rooms is rectangles. But It does not work well when the shape is either trapezium or L shape. Can somebody help me please?

Also it is not closing all the doors properly.

load('Walls')
e=ones(1,105);
A=imclose(imclose(~Image,e),e.');
B=imfill(A,'holes')&~A;
imshow(B)

[Matt J]

Will the doorways always be horizontal/vertical? And what example of an " L shape" didn't work? It seems to have worked well on the L-shaped flllor plan here:

https://www.mathworks.com/matlabcentral/answers/524224-how-to-get-the-coordinates-of-each-room-in-a-floor-plan-image-through-object-identification-method#answer_431573

[CHINTALA NIRMAL SRINIVAS]

L shape in the sense the shape of the room.In the example floorplan is L shaped but the rooms are rectangles. In the case where the rooms are either trapezium shaped or L shaped, the above code is not working properly

[Matt J]

And what about the doorways?

[CHINTALA NIRMAL SRINIVAS]

Not all the doors are closed but if I increase the number in ones() then they are being closed

Answer 1

[darova]

I skeletonized the image and found endpoints with bwmorph. Found shortest distances and connected points

I0 = imread('image.jpeg');
I1 = im2bw(I0);                     % binarize image
I2 = bwmorph(~I1,'skel',inf);       % make thin lines
I3 = bwmorph(I2,'spur',10);         % remove spurs
B  = bwmorph(I3,'endpoints');       % find free ends (endpoints)

[ii,jj] = find(B);                  % find rows and columns of endpoints
D = pdist2([ii(:) jj(:)],[ii(:) jj(:)]);    % find distances
D(D<1e-3) = nan;                    % remove zeros
[~,ix] = min(D);                    % find shortest distances

for i = 1:length(ix)
    i1 = ii(i):ii(ix(i));
    j1 = jj(i):jj(ix(i));
    I3(i1,j1) = 1;                  % connect door ends
end

I4 = imdilate(I3,ones(15));         % make lines thicker
imshowpair(I0,I4)                   % display original and modificated images

Result

[CHINTALA NIRMAL SRINIVAS]

does the regionprops() gives only rectangular bounding boxes?

[CHINTALA NIRMAL SRINIVAS]

Also it is not giving correct answer for other images.

Example:For this

The result I am getiing is

[darova]

• does the regionprops() gives only rectangular bounding boxes?

rectangular box can be only rectangular

• Also it is not giving correct answer for other images.

It's because there is no space between walls and margins

Try to add space

I0 = imread('image.jpeg');
[m,n,~] = size(I0);
I1 = ones(m+2,n+2);
I1(2:end-1,2:end-1) = im2bw(I0);    % binarize image

Are your images always have white background?

[CHINTALA NIRMAL SRINIVAS]

Yes they always have white backgrounds. How can I get the coordinates of each room if the bounding box is only rectangular in shape. I mean how can I get the coordinates of those rooms which are not in rectangular shape

[darova]

You can label image with bwlabel and select each region separately

[L,n] = bwlabel(I);
for i = 1:n
    I = L == i;
    imshow(I)
    pause(0.5)
end