Closed minnerbe closed 9 months ago
Currently, the Filter.process method only gets an ImageProcessor, so there is no way of matching the incoming image to the background in terms of stage coordinates.
Yeah - I guess to make this work, we'd have to inversely transform the full background image and chop it up into separate tiles. So, you'd end up with 4 background tiles for the thymus data set (2 rows x 2 columns). Not sure if this is doable.
Because of the success of https://github.com/saalfeldlab/render/pull/162 , @minnerbe thinks we should abandon this effort. We can keep this feature branch around for a while just in case we want to reference the changes. If/when we confirm it's not ever going to be needed, we can delete the branch and close this pull request.
I guess I'll go ahead and close this now ...
I started implementing background correction as discussed:
Filter
)However, I see some problems with this approach (technical and conceptional):
Filter.process
method only gets anImageProcessor
, so there is no way of matching the incoming image to the background in terms of stage coordinates.Alternative: we could use the already implemented
RollingBallSubtraction
filter that uses an ImageJ plugin to do local background correction for a single image. The results of applying this filter don't look too bad (original vs. rolling paraboloid vs. rolling ball):The downside would be that this is pretty computation heavy, so pre-applying the filter would be necessary, essentially duplicating the whole stack. What are your thoughts @StephanPreibisch ?