Non-transparent outside bounds with getTile for tifs without style

[banesullivan]

I noticed that TIF files with the TiffFileTileSource and OMETiffFileTileSource sources (at least), if the image is opened without a style, the areas outside the bounds of the image for getTile() are black/opaque but the encoding is PNG and these areas should be transparent

import large_image

path = 'test/test_files/test_orient0.tif'
source = large_image.open(
    path,
    encoding="PNG",
)

>>> img_bytes = source.getTile(0, 0, 0)
>>> img_bytes.mimetype
'image/png'
>>> img_bytes
<PNG below>

Or with an OME file:

path = 'HTA9_1_BA_F_ROI02.ome.tif'
source = large_image.open(
    path,
    encoding="PNG",
)
source.getTile(0, 0, 0)

However, with the OME source, if I open with a style, the region becomes transparent:

source = large_image.open(
    path,
    encoding="PNG",
    style = {"bands": [
      {"band": 1, "palette": ["#000", "#fff"]},
    ]}
)
source.getTile(0, 0, 0)

[banesullivan]

Nevermind. Adding a style does not solve it:

source.getTile(1, 1, 1)

[banesullivan]

Why is this an issue? It has significant impacts when using the new viewer in #1065:

Screen.Recording.2023-05-30.at.1.22.33.PM.mov

[manthey]

Does adding edge=#0000 (or any other transparent color) work?

Some viewers automatically crop to the maximum extent of the tiles; other viewers expect short tiles at the right and bottom edges (add edge=crop as a parameter). The edge color should only affect viewers that neither crop nor expect short tiles.

[banesullivan]

Setting an edge color changes it, but I still cannot get it to be transparent

Unfortunately, it looks like the ipyleaflet viewer cannot handle edge="crop"

edge="#00FFFFFF" and .getTile(1, 1, 1):

[manthey]

Do you mean edge=#FFFFFF00 (since the fourth octet is alpha)?

[banesullivan]

Yes I did mean that but still the problem I mentioned about not getting any transparency: Same tile with #FFFFFF00 (note white, not transparent border)

[banesullivan]

Turns out the issue is specific to greyscale images loaded without a style. If the images is greyscale, then the edge color remapped to the grayscale space. You have to apply a style to get out of the greyscale space:

import large_image

path = 'HTA9_1_BA_F_ROI02.ome.tif'

source = large_image.open(
    path,
    encoding="PNG",
    edge="#0F02",
)
source.getTile(1, 1, 1)

source = large_image.open(
    path,
    encoding="PNG",
    edge="#0F02",
    style = {"bands": [
      {"band": 1, "palette": ["#000", "#fff"]}
    ]}
)
source.getTile(1, 1, 1)

This is not a high priority