Feature request: display() should be able to add images to an existing viewer

[kevinyamauchi]

I think it would be useful if display() had the option to add the image as a new layer to an existing Napari viewer. This is super useful for comparing the image before/after a processing step (e.g., filtering).

Thoughts, @kne42 and @ambrosejcarr ?

[kevinyamauchi]

I am happy to take a stab at this if you all agree that it's useful.

[ambrosejcarr]

This is super useful for comparing the image before/after a processing step (e.g., filtering).

Could you help me understand what advantage there is to having the images be on the same viewer vs just tabbing back and forth between different viewers? Are you using transparency somehow?

[kevinyamauchi]

I think the main reason including multiple ImageStack in the same viewer is useful is that it keeps them aligned. That way as you pan and zoom between different features, you can immediately compare how they are changed between different processing steps.

In terms of how to compare, I sometimes use opacity as you mentioned, but I mostly hide/show specific layers (the eyeball on the right hand side tool bar).

[ambrosejcarr]

I think the main reason including multiple ImageStack in the same viewer is useful is that it keeps them aligned. That way as you pan and zoom between different features, you can immediately compare how they are changed between different processing steps.

Got it, makes sense to me! It'd be great if you would take a stab at this. How do you imagine changing the interface to support a persistent viewer?

[kevinyamauchi]

I propose to add an optional input argument that takes a napari Viewer object. If a Viewer object is provided, display() would append the image/spots as new layers in the provided Viewer.

The display() returns the Viewer object, so the usage pattern would be:

# Look at the unprocessed image
viewer = display(stack=unprocessed_stack)

# Do some filtering
ghp = Filter.GaussianHighPass(sigma=3)
high_passed = ghp.run(primary_image, verbose=True, in_place=False)

# Admire your results
display(stack=high_passed, viewer=viewer)

A question I have: my understanding is that the Viewer object would be passed by reference display(). When the user passes in a Viewer to append to, should display() return the reference to the Viewer object or nothing?

[kne42]

Sounds like a great API to me. It should return the same viewer as passed in :)