Height, width not set in level 0

[jonasteuwen]

In the function isyntax_init_level_offsets, I believe in level 0 we also need to set width and height, so perhaps we can set them (otherwise libisyntax_level_get_width and libisyntax_level_get_height will return 0.

In my experiment with an isyntax file, it looked like that there is an offset in level 0 as well.

[Falcury]

Ok, I’ll look into this tomorrow.

[Falcury]

I fixed this in b837f88.
I also moved the initialization of these values to a more appropriate place in isyntax_open().

[jonasteuwen]

Great! I will verify with a few samples, and also check when written to .tiff if it matches the SDK pixels.

[Falcury]

I've noticed in the past that TIFF files converted from iSyntax using the SDK (not IMS export) have their offsets shifting between levels. So that might also be a source of problems.

To make the shift between levels more easily visible I added a debug slider to Slidescape, see the screenshot below:

You could also use this to verify that the level offsets we are calculating for iSyntax are (visually) correct.

[jonasteuwen]

That's great! Will use it.

[jonasteuwen]

@Falcury It's very useful thanks!

If I use this and convert the tiff according to my read_region, it's very jumpy and it doesn't look like the offset is correct in origin_offset_in_pixels. If I use int32_t offset = ((PER_LEVEL_PADDING << level_count) - PER_LEVEL_PADDING) >> level;, this does not (really) happen, but then there appears to be an absolute offset compared with the Philips SDK.

There is a very minor shift when you use the offset I'm giving, but for the currently defined one from level 3 onwards it really shifts

[jonasteuwen]

Check #17, I think the formula given by me is correct

[jonasteuwen]

Seems to be addressed.

[Falcury]

I had to revert to storing the height/width of the WSI including the padding, because this affects the calculation of the grid width and height. This was causing some codeblocks to not get initialized, leading to a crash on some iSyntax files (specifically, the example image 3.isyntax that was provided by Philips on their original OpenPathology portal).
I added and renamed new width/height fields with the padding subtracted.
Commit: 5acc98b

[jonasteuwen]

Great catch! Aren’t the names not becoming a bit…ugly?

[Falcury]

Great catch! Aren’t the names not becoming a bit…ugly?

Well, yes they are, but I figured I had to make a descriptive distinction somehow ;-) How would you name them?
I figured it would be OK to rename these values to something else, also because they're hidden behind a getter in the public API...

[jonasteuwen]

Perhaps, since they actually are the height and width of the image itself, it could be level_height, or height and rename the others?