How can i get the smallest shape using a list of centroid coordinates?

As part of image analysis protocol, I have binary images with shapes I have detected using region props. As the dimensions of the original area containing the shapes differs outside the edges of the shape-enriched area, I am trying to find the smallest shape that includes all the dots (shapes detected with regionprops). The approaches I have looked for deal with paths that intersect with all the shapes, hamiltonian paths and whatsoever, but I only need the outer shapes to be the actual vertexes of the shape, the ones inside the resulting shape need to be left excluded.

If you need a visual explanation, imagine a circle-like distribution of stars, how can I get a shape that includes all the stars, both the ones that end up making the shape perimeter and the ones that are left inside?

Thank you!

[Jonas]

is the distribution of your stars concave? than maybe a convex hull using bwconvhull may be what you are searching for?

if I misunderstood, could you show an image and mark what you want to get from it?

[Miguel Ramirez]

Its wedged shape (it is a biological sample so it is subjetcted to change). I just compared your suggestion(white shape) with the alphashape i obtained. It looks the latter is much more conservative with the actual shape (sickle versus wedge).

Answer 1

[Matt J]

You can use bwalphaclose from this FEX download,

https://www.mathworks.com/matlabcentral/fileexchange/90762-further-tools-for-analyzing-objects-in-n-dimensional-images?s_tid=srchtitle

It is very similar in its syntax to bwconvhull, except you can specify the alphashape you have already designed to take a more form-fitting closure.

[Miguel Ramirez]

It looks like basic area(alphashape) is good enough for this, but i will inspect your tools to see if it gets improved. thanks!

Answer 2

[Image Analyst]

Why don't you just simply OR all the binary images together?