help me get the points of the mask

I have shapes like this and i want the four points like what i have marked and the enclosing are i drew with lines

[Matt J]

Please provide us the raw imges (in .mat attachments) so we can demonstrate solutions.

[Matt J]

Your decompositions are ambiguous. Why in some cases do the quadrilaterals overlap, but not in others?

[kunal]

here it is:-

just use and run

% Load all masks from the .mat file
load('my_masks.mat');

figure;

subplot(2,2,1);
imshow(a1);
title('Mask 1');

subplot(2,2,2);
imshow(a2);
title('Mask 2');

subplot(2,2,3);
imshow(a3);
title('Mask 3');

subplot(2,2,4);
imshow(a4);
title('Mask 4');

[Matt J]

Thanks, but you still haven't addressed my second comment. The partitioning you've drawn on the figures is overlapping in some cases and, for some unclear reason, non-overlapping in others.

[kunal]

ah yes, it's ok if it overlaps or not, sorry if this is confusing. As you can see every part or side i have enclosed it in perfect square/rectangle or four point polygon and where the mask is continuous it also shares 2 point of edges

[Image Analyst]

@kunal did you not understand my Answer below or did you not even see it (scroll down)? I didn't see any response whatsoever to it.

Answer 1

[Image Analyst]

You can find a wealth of information online just by Googling "Minimum perimeter polygon". Yes, it's a thing. And there is much research and published papers and algorithms on it.

Google "Minimum Perimeter Polygon"

"In digital image processing, a minimum perimeter polygon (MPP) is a technique to represent the boundary of a shape or object using a polygon with the fewest possible edges while maintaining the essential shape characteristics. It's a form of polygonal approximation where the goal is to approximate a digital boundary with a polygon with the smallest perimeter."

Also see the PDF I've attached.

Sorry, I don't have any MATLAB code for it but you could try asking MATLAB's AI engine to write it for you:

MATLAB AI Chat Playground

You might also look at the LIDAR toolbox : https://www.mathworks.com/help/lidar/

See examples at:

https://www.mathworks.com/help/vision/examples.html?category=point-cloud-processing

[kunal]

this consider the image as an whole but i want to seperate the pieces into part of rectangle

[Image Analyst]

Each "blob" in the image is considered a separate, distinct polygon. You can get the minimum points to define each using the concepts of minimum perimeter polygon.

If you have a single blob that is shaped like an L, the min perimeter polygon will define the whole L shape but it will not split the two L into separate blobs. Not sure why you'd really need that though instead of just a single polygon.

Answer 2

[Matt J]

You can try bwlpolyshape(), downloadable from,

https://www.mathworks.com/matlabcentral/fileexchange/181422-fit-polyshape-to-objects-in-image

which will fit a polyshape to the mask, subject to a distance tolerance distTol. You will have to do some postprocessing to consolidate some of the spurious vertices.

S=structfun(@logical,load('my_masks'),'uni',0);

p=bwlpolyshape(S.a2,distTol=15,refine=0,Visualize=true)

p = 
  polyshape with properties:

      Vertices: [20×2 double]
    NumRegions: 1
      NumHoles: 2

[x,y]=boundary(p);

hold on
plot(x,y,'og','MarkerFaceColor','y');hold off

●