Comments on: Multidimensional Image Processing

By: Steve

Steve — Mon, 01 Nov 2010 11:56:19 +0000

Roman—Thank for your input on uses for the largest incribed box.

By: Roman

Roman — Fri, 29 Oct 2010 05:55:05 +0000

i’ve seen files on matlab central that will fit the smallest triangle/circle/rectangle about a group of objects or points, but they are always set mathematical shapes. whereas, if i was going to make a new mask for my thrombus, i need a free shape of minimal area, that will envelope all of my bw objects… and that seems like a project within itself. so the easier solution to do the maximum inscribed box inside of the not so accurate mask that i already have.

By: Roman

Roman — Fri, 29 Oct 2010 05:41:30 +0000

if you mean an application for the *largest* inscribed box, yes I have several:
1) when i was working on my phd, i received Xray scans of a sample that was contained within a circular holder. i needed to do a flow dynamics simulation through the geometry obtained from the xray. my simulation uses periodic boundary conditions, and i wanted to model an “infinite” sample in an directions. so pretty much I needed to cut out the largest usable rectangular piece of the image that was contained within the circular holder… in that case the problem was simpler, because you could solve it analytically, and then just fit the circle to the holder, find its radius and apply the formula

2)my project now, as i said before, involves imaging a thrombus that is forming a blood vessel of a mouse. so after i do my image filtering and segmentation, i am left with an odd shaped porous object. i am interested in measuring its porosity. the mask that i have for the *whole* object is a signal that comes from a florescent marker in the blood, but the problem is that it encompasses a pocket of fluid under the thrombus (which i don’t want to include in my calculation of porosity).

so my solution is again, fitting a largest box that fits inside of the bw objects (cells), and measuring the porosity inside of that box. maybe there is some way to isolate to draw a perimeter around all of my bw objects, and then filling in the holes to make a new mask, but i don’t know how to do the former part.

in any case, later on in my project, if i want to do flow through an ‘infinite’ thrombus, to get its permeability for example, i will again need the largest 3d rectangular cutout from inside of the thrombus. so the largest inscribed box is very useful for people like me, who use the results of image processing for simulation.

By: Roman

Roman — Thu, 28 Oct 2010 18:31:56 +0000

Steve,

I also wanted to ask you, is there some good way to ensure 3D connectivity between my objects? For example, my 3D BW matrix is a collection of cells that make up a thrombus. I don’t want anything just floating in free space, so I would like to impose a constraint that everything must be connected (no disconnected objects). Any advice? Thanks

By: Steve

Steve — Thu, 28 Oct 2010 14:00:41 +0000

Roman—Thanks for the suggestion. Do you have a specific use case in mind for the smallest incribed box measurement?

By: Roman

Roman — Thu, 28 Oct 2010 06:28:19 +0000

thanks… can i also suggest a largest inscribed box in addition to the smallest inscribing box, to the regionprops function? seems like another basic and useful thing that is missing.

By: Steve

Steve — Wed, 27 Oct 2010 20:01:35 +0000

Roman—bwmorph is a very old function that dates back to version 1 in 1993. There are aspects of its original interface (parts of which are no longer documented) and implementation (heavily lookup-table based) that force 2-D operation. It also supports operations that aren’t well defined in 3-D. (I would include ‘spur’ in that set.)

I think we should take a fresh look at this function and modernize it, possibly by providing parts of its functionality as new functions that support multidimensional operation.

Regarding your desire to use ‘clean’ in three dimensions, you can get this effect by calling bwareaopen.

By: Roman

Roman — Wed, 27 Oct 2010 18:42:31 +0000

Steve, two years after Mark’s post about bwmorph in 3D I am also wondering the same thing… I have a 3D matrix of BW objects that I would like to clean up using ‘clean’ and ‘spur’ in 3D. I guess it wouldn’t be too hard to write this myself by labeling objects and counting how many voxels belong to each one, but it seems so basic that I was expecting Matlab to have it.

By: Steve

Steve — Wed, 23 Apr 2008 11:53:33 +0000

Mark—There is a 3-D skeletonization routine available on the MATLAB Central File Exchange you could try. But now I see from the reviews of that submission that you already found it. I would be very interested to know if the method works for your problem.

By: Steve

Steve — Wed, 23 Apr 2008 10:35:54 +0000

Mark—I don't know. It won't happen for R2008b. imdilate and imerode do work in three (or higher) dimensions, but they implement dilation and erosion, which are different from the 2-D 'thicken' and 'thin' operations offered by bwmorph.