Support for slicing and concatenting?

[dickinson-lab]

First of all, many thanks for this extremely useful tool - it's been a game-changer for me when working with multi-gigabyte tiff files in MATLAB.

I was wondering if you'd consider adding functions to slice and concatenate images without loading the entire file. As a trivial example, suppose I have a TIFFStack object with size [600, 1200, 500] and I want to pass just half of it to another function. I would be tempted to do something like
smallTIFFStack = bigTIFFStack(:,1:600,:);
but this loads everything into memory, when what I want is a new TIFFStack object that just references the portion of the image I selected.

It also would be nice to be able to concatenate two TIFFStack objects, for cases when my image acquisition software has split a very long timeseries into multiple files:
`>> size(bigTIFFStack1)
ans = 600 1200 4000

size(bigTIFFStack2)
ans = 600 1200 3745
evenBiggerTIFFStack = cat(3, bigTIFFStack1, bigTIFFStack2;
size(evenBiggerTIFFStack)
ans = 600 1200 7745`

If you don't have time to do this, I'd be willing to work on the coding, but I'd need some help figuring out where to start. I am proficient with MATLAB but don't really know anything about how it deals with memory-mapped files.

[DylanMuir]

Hi @dannyhmg, thanks for your comments! So for concatenation there is https://github.com/DylanMuir/TensorStack which should work out of the box for what you want. Slicing is much harder, but I had a vague idea that this was already supported somewhere — give me a moment to look.

I haven’t touched this code for quite a while, though. And supporting tensor referencing so it looks like a MATLAB tensor is really complicated!

[DylanMuir]

Ah, I was thinking of the de-interleaving in TIFFSTack. So there's no existing functionality to do this. I would suggest making a small wrapper class that does slicing on a tensor, and just passes the indices through. Maybe take a look at https://github.com/DylanMuir/LazyTensorTransform

[dickinson-lab]

Hi Dylan, thanks very much for taking the time to reply.

I tested TensorStack and it indeed does exactly what I want - thank you!. I noticed that unlike TIFFStack, it doesn't include a license file. May I use a clone of TensorStack as part of an open-source package I am developing (https://github.com/dannyhmg/SiMPull-Analysis-Software), and if so, would you like to add a license?

Slicing is less urgent since I realized that in lieu of slicing a TIFFStack object and passing the result to a function, I can just pass indices along with the whole object along, and then let my function load the part of the image it needs. I might revisit writing a wrapper class later if passing indices becomes problematic.

[DylanMuir]

Great! I’m happy about that. You’re welcome to use TensorStack in that way. I’ll add the same license as TIFFStack.

[dickinson-lab]

Great - thanks again for your help and for sharing your work.