Hi Dat, Your are right on the field of view (FOV) score, one requires the screen distance, height and width to compute the vertical and horizontal field of view.
The fushion distance/parallax distance is not appropriate though. The concept of what you are focussing on the scene completely orthogonal to the display system set up. Furthermore any rotation of the view matrix for each each is not related to screen distance, but purely the setup of displays relative to the viewers eye, this is person/HMD hardware configuration attribute, and is very different from that attributes of a projected/monitor based stereo display. I would also strongly recommend not thinking about convergence angles, as this really stems for the early mistaken attempts at doing stereo that sadly got published as an easy way of adding stereo to your application. Translating then rotation the view matrix is not the same mathematically as translating the view matrix and shearing the projection matrix which is the right way to do projected/monitor based stereo, and as a consequence gives you incorrect stereo effect. You seem to have picked up on this broken way of thinking about stereo, and then further trying to apply it to HMD based stereo which doesn't need any rotation save for accounting for the physical configuration of the displays. Robert. On Sun, Dec 21, 2008 at 1:56 AM, tien dat <[email protected]> wrote: > Hi Robert, > This might go beyond my original question, but I just want to share my > thought about your point. > First of all, I think screen distance does have some effect here, > since the farther the screen, the smaller the FOV (people have > different head shapes, right?). > Second, I think iod is not the only one parameter that you concern > about. Convergence angle (i.e. fusion distance) should also be one > important factor because this is where you really "focus" at. I don't > know how much different if you set this point to infinity instead of > some objects that you should focus on. I don't know the answer, so I > want to try to see by myself, and have the beginning question. > Anyway, thank you very much for your help. I will ask you when I have > more question about OSG. > Dat > > On 12/20/08, Robert Osfield <[email protected]> wrote: >> HI Dat, >> >> I didn't implement this specific head mounted elements, it's what an >> end users submitted for his setup. I don't have any head mounted >> displays to tests against. This code is also pretty old and not part >> of the standard stereo setup. >> >> I do however I lot of experience with driving projection based stereo >> systems, head mounted displays aren't a far stretch from this but >> there are differences as I said in my previous email. Personally I'd >> set up the slave cameras myself and totally ignore the screen distance >> as it is has no relevance to HMD's. The only thing of relevance is >> that inter-occulare distance which requires a horziontal translation >> of the view matrix. >> >> Robert. >> >> On Sat, Dec 20, 2008 at 6:43 PM, tien dat <[email protected]> wrote: >>> Hi Robert, >>> I don't totally agree with your answer. In these two functions >>> computeRightEyeViewImplementation() and >>> computeLeftEyeViewImplementation(), you do use iod, screen distance >>> and fusion distance in computing the view matrices. And you don't >>> check to see if the display mode is HEAD_MOUNTED_DISPLAY or not - as >>> you do in computing the projection matrices. So I still think that you >>> do use these variables in computing the view matrices when the display >>> mode is HMD. >>> Could you give me a little more explanation for that. >>> Thanks, >>> Dat >>> >>> On 12/20/08, Robert Osfield <[email protected]> wrote: >>>> Hi Dat, >>>> >>>> The screen distance and fusion distance are relevant to monitor or >>>> projection based stereo systems, not for HMD's. The OSG doesn't use a >>>> convergence angle as this is not appropriate for proper stereo effect >>>> in monitor or projection based stereo system, instead the OSG uses the >>>> mathematically correct system of shearing the projection matrix. >>>> >>>> Things a bit different for HMD, as the projection matrix is identical >>>> for both eyes, and only the view matrix is offset (and perhaps >>>> rotated). One method to accurate control this is to use a >>>> SceneView::ComputeStereoMatricesCallback, but now with have the >>>> osgViewer it's probably easier just to manually create two slave >>>> Camera's with the appropriate view matrix translation to account for >>>> the inter-ocular distance. >>>> >>>> FYI, osgViewer hasn't yet taken over management of all stereo modes >>>> from the lower level SceneView, but this is my plan - to migrate all >>>> stereo support into the high level class, and then provide >>>> configuration file support for it. This will be done well after 2.8. >>>> >>>> Robert. >>>> >>>> On Sat, Dec 20, 2008 at 3:37 AM, tien dat <[email protected]> wrote: >>>>> Hi all, >>>>> >>>>> I have a question about controlling stereo with HMD in OpenSceneGraph. >>>>> >>>>> Basically to setup stereo display you need the distance of two eyes >>>>> and the convergence angle. When I take a look at OpenSceneGraph code >>>>> for setting up stereo with HMD, I see two other variables which are >>>>> screen distance and fusion distance. There is also a matrix to >>>>> multiple with each view for each eye that uses these variables. So >>>>> could any of you tell me what they mean and what happens if I change >>>>> these variables. >>>>> >>>>> Thank you very much, >>>>> >>>>> Dat >>>>> _______________________________________________ >>>>> osg-users mailing list >>>>> [email protected] >>>>> http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org >>>>> >>>> _______________________________________________ >>>> osg-users mailing list >>>> [email protected] >>>> http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org >>>> >>> >>> >>> -- >>> ------------------------ >>> Dat Tien Nguyen >>> Email: [email protected] >>> Address: 218/74/7 Tay Son Dong Da Hanoi Vietnam >>> Phone: 844-5370595 >>> _______________________________________________ >>> osg-users mailing list >>> [email protected] >>> http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org >>> >> _______________________________________________ >> osg-users mailing list >> [email protected] >> http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org >> > > > -- > ------------------------ > Dat Tien Nguyen > Email: [email protected] > Address: 218/74/7 Tay Son Dong Da Hanoi Vietnam > Phone: 844-5370595 > _______________________________________________ > osg-users mailing list > [email protected] > http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org > _______________________________________________ osg-users mailing list [email protected] http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
