Gaussian blur depth cueing? (ChimeraX)
Hi, I was wondering if you have thought about implementing depth-dependent gaussian blur as a depth cueing mechanism in ChimeraX? The idea would be to blur the image in a manner dependent on the distance from the plane intersecting with the center of rotation, to mimick depth of field in a camera. Mocked up in the attached using a tilt/shift effect, but without access to real depth information - I wonder if it might better convey depth information than distance-dependent changes in brightness/contrast under some circumstances? Cheers Oli
Sort of related to this - any chance for symmetric depth cueing as an option in ChimeraX? So that instead of intensity falloff being linear from the front clip plane to the back clip plane, it is triangular - intensity linearly increasing from the front clip plane to the midpoint, then decreasing towards the back clip plane? This would be helpful when making closeup clipped images of models with maps, where fragments of regions in front of the region of interest can obscure the intended subject, which is somewhat distracting. Cheers Oli
On May 5, 2017, at 9:42 AM, Oliver Clarke <olibclarke@gmail.com> wrote:
Hi,
I was wondering if you have thought about implementing depth-dependent gaussian blur as a depth cueing mechanism in ChimeraX?
The idea would be to blur the image in a manner dependent on the distance from the plane intersecting with the center of rotation, to mimick depth of field in a camera.
Mocked up in the attached using a tilt/shift effect, but without access to real depth information - I wonder if it might better convey depth information than distance-dependent changes in brightness/contrast under some circumstances?
Cheers Oli
<test_image_tilt_shift.png>
Hi Oliver, Usually clip planes are used to hide parts of maps and atomic models in front of and behind a slab of interest to get a clear view of the slab. If the slab is too thick, then too many parts of the structures are on top of each other and it is hard to see things. Depth cueing can dim the back so a slightly thicker slab can be shown without cluttering the view the stuff in front. I know you know all this. Having decreased brightness in front does not seem useful to me — it makes dim stuff in front block the view of the bright stuff behind it. I see that it can provide context, letting you know what structures are just in front of the bright region of interest, but I doubt it is a good idea to obscure the bright region where attention is focused. That said, if you are looking at a map mesh, the dimming in back and dimming in front aren’t really any different in the degree they obscure the bright middle region. This points to another problem that dim stuff in front will be visually confusing — will require rotation motion to figure out what dim parts are really in front. Like the depth blurring idea, it is easy to make a mockup image. In Chimera you can make the front dim and back bright by making the depth cue start fraction larger than the depth cue end fraction (menu Tools / Viewing Controls / Effects). Then make another image with normal depth cuing and blend the two images in a photo editor (using a min intensity combiner). I’m looking at the interactive reversed depth cue now in Chimera and it is really unpleasant, especially when rotating because the depth looks inverted. Again I think a compelling mockup image is the first step before investing time trying to implement this. The “reverse depth cue” trick does a pretty crappy job because it fades the mesh lines to the background color — it does not make them transparent, so they strangely block the brighter mesh lines in back where the cross. To make it work correctly with transparency would be either difficult or slow because multilayer transparency not easy to handle in OpenGL. All the more reason that a compelling mockup would be important to justify spending time coding such an effect. Tom
On May 5, 2017, at 11:59 AM, Oliver Clarke <olibclarke@gmail.com> wrote:
Sort of related to this - any chance for symmetric depth cueing as an option in ChimeraX?
So that instead of intensity falloff being linear from the front clip plane to the back clip plane, it is triangular - intensity linearly increasing from the front clip plane to the midpoint, then decreasing towards the back clip plane?
This would be helpful when making closeup clipped images of models with maps, where fragments of regions in front of the region of interest can obscure the intended subject, which is somewhat distracting.
Cheers Oli
On May 5, 2017, at 9:42 AM, Oliver Clarke <olibclarke@gmail.com> wrote:
Hi,
I was wondering if you have thought about implementing depth-dependent gaussian blur as a depth cueing mechanism in ChimeraX?
The idea would be to blur the image in a manner dependent on the distance from the plane intersecting with the center of rotation, to mimick depth of field in a camera.
Mocked up in the attached using a tilt/shift effect, but without access to real depth information - I wonder if it might better convey depth information than distance-dependent changes in brightness/contrast under some circumstances?
Cheers Oli
<test_image_tilt_shift.png>
_______________________________________________ Chimera-users mailing list: Chimera-users@cgl.ucsf.edu Manage subscription: http://plato.cgl.ucsf.edu/mailman/listinfo/chimera-users
Hi Oliver, The depth of field blur images I have seen of maps and molecules have all struck me as cool looking and less useful then conventional brightness based depth cuing. Brightness depth cue allows you to include more of the structure in back while reducing the obscuring effect it has on stuff in front. Blurring depth cue interferes much more with the stuff you want to be in focus. Your image minimizes the badness by not having any blurring in the central region of the image, but the periphery where there is blurring may be artistically nice, but I hope you are not looking at the structure there. That said the technique could be done using the depth buffer that assigns depth to each pixel — it can’t be done really well that way because that depth only gives the front-most object and if transparency is present the pixel shows multiple objects at different depths. I think it would require a compelling mock-up image to make it worth spending effort on this effect. It is relatively easy to make a mock-up image by using clip planes to save several images representing slabs at different depths, then blur each slab image differently depending on depth in a photo editor and composite. Tom
On May 5, 2017, at 6:42 AM, Oliver Clarke <olibclarke@gmail.com> wrote:
Hi,
I was wondering if you have thought about implementing depth-dependent gaussian blur as a depth cueing mechanism in ChimeraX?
The idea would be to blur the image in a manner dependent on the distance from the plane intersecting with the center of rotation, to mimick depth of field in a camera.
Mocked up in the attached using a tilt/shift effect, but without access to real depth information - I wonder if it might better convey depth information than distance-dependent changes in brightness/contrast under some circumstances?
Cheers Oli
<test_image_tilt_shift.png> _______________________________________________ Chimera-users mailing list: Chimera-users@cgl.ucsf.edu Manage subscription: http://plato.cgl.ucsf.edu/mailman/listinfo/chimera-users
Hi Tom, here is a more realistic example - I have taken a stack of overlapping 2Å thick density/model slabs in Chimera, no depth cueing applied. I then applied depth cueing either using brightness/contrast, blur, or both (in Affinity photo). I haven’t taken much care to optimize the parameters in either case, but I do think the blur provides at least was good a depth perception effect as brightness/contrast, and better when viewed at a larger size. Cheers Oli
On May 8, 2017, at 10:07 PM, Tom Goddard <goddard@sonic.net <mailto:goddard@sonic.net>> wrote:
Hi Oliver,
The depth of field blur images I have seen of maps and molecules have all struck me as cool looking and less useful then conventional brightness based depth cuing. Brightness depth cue allows you to include more of the structure in back while reducing the obscuring effect it has on stuff in front. Blurring depth cue interferes much more with the stuff you want to be in focus. Your image minimizes the badness by not having any blurring in the central region of the image, but the periphery where there is blurring may be artistically nice, but I hope you are not looking at the structure there.
That said the technique could be done using the depth buffer that assigns depth to each pixel — it can’t be done really well that way because that depth only gives the front-most object and if transparency is present the pixel shows multiple objects at different depths. I think it would require a compelling mock-up image to make it worth spending effort on this effect. It is relatively easy to make a mock-up image by using clip planes to save several images representing slabs at different depths, then blur each slab image differently depending on depth in a photo editor and composite.
Tom
On May 5, 2017, at 6:42 AM, Oliver Clarke <olibclarke@gmail.com <mailto:olibclarke@gmail.com>> wrote:
Hi,
I was wondering if you have thought about implementing depth-dependent gaussian blur as a depth cueing mechanism in ChimeraX?
The idea would be to blur the image in a manner dependent on the distance from the plane intersecting with the center of rotation, to mimick depth of field in a camera.
Mocked up in the attached using a tilt/shift effect, but without access to real depth information - I wonder if it might better convey depth information than distance-dependent changes in brightness/contrast under some circumstances?
Cheers Oli
<test_image_tilt_shift.png> _______________________________________________ Chimera-users mailing list: Chimera-users@cgl.ucsf.edu <mailto:Chimera-users@cgl.ucsf.edu> Manage subscription: http://plato.cgl.ucsf.edu/mailman/listinfo/chimera-users <http://plato.cgl.ucsf.edu/mailman/listinfo/chimera-users>
participants (2)
-
Oliver Clarke -
Tom Goddard