Describe the bug
TissueMask produces different results, depending on whether or not large_image is used when reading the slide. I have tested this behaviour with 2 different slide formats (.mrxs and .svs).

To Reproduce
See attached screenshot.

Expected behavior
The mask returned by TissueMask should be the same (or very close), independent of the backend used for reading the slides. This should be the case for all methods.

Screenshots

Software (please complete the following information):

• OS: Linux
• Python Version: 3.9
• histolab version: 0.4.1
• large_image: 1.11.1
• pixman: 0.40.0

Additional context
Several other issues were triggered depending on the backend:

• When opening an .MRXS file, the background of the slides is different depending on whether large_image is used or not. In the former case, the background is black, while in the latter is white. This is probably due to the way OpenSlide and large_image handle the alpha-channel from the .MRXS: using it as a mask to set background to white or simply ignoring it.
• (as a consequence of the above) TissueMask returns completely different results in the two use cases.
• In the case of the .SVS file, the thumbnail image has different size depending on which backend was used.