dials.image_viewer: Display -2 as "masked" occurs for non-Pilatus images

[dagewa]

An example is in the backstop region of this image:

dials.image_viewer $(libtbx.find_in_repositories dials_regression)/image_examples/ESRF_BM14/mar225_1_001.mccd

For those with access, the images in /dls/mx-scratch/rjgildea/fisher/013_021_G4 are worse, with many "masked" pixels across the image

[rjgildea]

This is a known thing, and I'm not sure a fix is possible without a significant reworking of how the display is handled in the FlexImage class:
https://github.com/cctbx/cctbx_project/blob/master/iotbx/detectors/display.h#L236

[dagewa]

I wonder if @nksauter would like to comment? Why do we need special code to display -2 as masked anyway? Shouldn't the masked pixels be defined by the mask?

[rjgildea]

The code in FlexImage long-predates dxtbx and image masks (and also Pilatus detectors) - judging by labelit svn logs it dates back to 2004.

[dagewa]

I haven't yet tried it, but I thought for a simple fix that this line: https://github.com/cctbx/cctbx_project/blob/master/iotbx/detectors/display.h#L199
should instead be
bool has_pilatus_inactive_flag = false;
and then it is set true individually for DECTRIS detectors based on vendortype and then doesn't end up true for e.g. CCD datasets

[rjgildea]

Yes, but then it won't be possible to view image masks for CCD images. The problem is that the image viewer is (ab)using the functionality in FlexImage flagging the special Pilatus -2 pixels in order to display dxtbx image masks.

[graeme-winter]

I would definitely say that @nksauter would hold an opinion here as clearly this must also be a problem for pedestal subtracted CSPAD images…. Properly supporting the concept of a mask not derived from pixel values seems like a good idea…

[rjgildea]

dials/util/image_viewer/spotfinder_frame.py

Lines 1077 to 1080 in 42fcf5f

	def mask_raw_data(self, raw_data):
	mask = self.get_mask(self.pyslip.tiles.raw_image)
	for rd, m in zip(raw_data, mask):
	rd.set_selected(~m, -2)

[dagewa]

I made the change anyway to see what happens. I defined a rectangular mask, imported with that then viewed in the image viewer. When "Show mask" is selected the -2 valued pixels are no longer red, and the masked region becomes white.

[dagewa]

Original code gets the red coloured mask, which is obviously better, but has the spurious -2 pixels "masked" as well.

[rjgildea]

Presumably they are only appearing as white because of the code above that sets the masked pixels to -2, and -2 is a "small number", so the display is set to white (or close to white) for those pixels (with that colour scheme).

[nksauter]

I'd like the opportunity to work on this problem, so that we can account for both the Pilatus inactive pixels, and the possibility of displaying masks separately. Realistically my first opportunity to even think about the problem would be during the week of May 6. If we can hold the issue open till then I'll have a look. Nick Nicholas K. Sauter, Ph. D. Senior Scientist, Molecular Biophysics & Integrated Bioimaging Division Lawrence Berkeley National Laboratory 1 Cyclotron Rd., Bldg. 33R0345 Berkeley, CA 94720 (510) 486-5713

…

On Mon, Apr 16, 2018 at 9:47 AM, Richard Gildea ***@***.***> wrote: Presumably they are only appearing as white because of the code above that sets the masked pixels to -2, and -2 is a "small number", so the display is set to white (or close to white) for those pixels (with that colour scheme). — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#536 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AOGuVqlRYGef44O-D3t1UV_7XorsvZ0kks5tpMsXgaJpZM4TVI15> .

[dagewa]

Hi @nksauter, I don't think any of us are going to touch this until you have a chance to take a look. Thanks.

[dagewa]

See also cctbx/cctbx_project#49

[dagewa]

Just noting that I get problems with this particularly with electron diffraction datasets where there may be genuine negative-valued pixels. I'd like to see what the values are, but of course the image viewer always tells me they are -2.