Support compressed tifs for read/write

[freeman-lab]

In the refactor currently underway (#237) we'll add a totif method that saves an Images object to tif files. @d-v-b can you post here the code snippet that adds optional compression when using sklearn.external.tifffile?

[d-v-b]

compression is passed as an integer kwarg to the imsave() function, e.g. imsave(im_path, image, compress=1). The default, compress = 0 means no compression, and increasing values of compress above 1 increase the level of compression, which I think is done via zlib.

More generally, there are other kwargs that people might want to use when saving tifs, since there's plenty of metadata that can be set in the tif format. How possible is it to completely wrap the functionality of tiffffile in thunder?

[freeman-lab]

Ok great thanks @d-v-b ! Generally it'd be easy to add whatever kwargs we want, because the thunder.images totif method is more or less just calling the tifffile imsave method on every record. One option is to list them and pass them along explicitly, the other is to just to pass all extra kwargs from totif to imave . I sort of prefer the former because it can be more explicitly documented.

[d-v-b]

Agreed @freeman-lab. What do you think about more complex cases of thunder wrapping some function from another module that has kwargs? timeSeries.normalize(baseline=window-fast) comes to mind, where scipy.ndimage.filters.percentile_filter is called with mode = nearest, but the user might want something else... (mode = nearest causes strong edge effects). How should this kind of case be handled?