[x3d-public] OrthoViewpoint fieldOfView

[Michalis Kamburelis]

2018-05-04 22:57 GMT+02:00 Andreas Plesch <andreasplesch at gmail.com>:
> I think it is intended as an implementation note to suggest the
> maximum/minimum_x/y need to be recomputed by the browser to fit the
> display aspect ratio before being used in a projection matrix but I am
> not sure. I may also be a warning that the fieldOfView parameters need
> to have the same aspect ratio than the viewport which, however, is
> typically not known to the scene author in advance. Overall, the note
> may be more confusing than helpful.

Indeed (as far as I understood and implemented), the
"OrthoViewpoint.fieldOfView" must be adjusted to match the aspect
ratio of the window, when being used for projection. The specified
"OrthoViewpoint.fieldOfView" in X3D file describes only the *minimum*
view. So the X3D author controls precisely what *must* be visible in
the window, and leaves up to the renderer to eventually show a little
more, depending on window aspect ratio.

This is consistent with perspective "Viewpoint.fieldOfView" treatment.
It also describes a *minimum* angle.

BTW, I just noticed there's a simple bug in this specification text:
the text about "X3DViewpointNode" also talks about "fieldOfView", "The
fieldOfView field specifies a preferred minimum viewing angle....".
Which is incorrect, as "X3DViewpointNode" does not define
"fieldOfView". Only the "Viewpoint" and "OrthoViewpoint" define
"fieldOfView" (with different types, but consistent treatment).


> There is also the question why the four corners should be provided and
> not just a width and height, eg. two parameters, more similar to the
> perspective field of view. In theory, providing the four corners
> allows for an asymmetric projection, eg. an additional shift away from
> the center. Since the center is also the center of rotation of examine
> mode, this option seems exotic and not very useful. In fact, x3dom
> seems to override any asymmetry and just uses the width and height
> (deltas) with 0 offset. It looks x_ite allows that, on the other hand,
> although I did not test. There was perhaps some additional
> consideration when the field was introduced ? Does anybody recall ?

One can change the center of rotations using the centerOfRotation
field (available on both OrthoViewpoint and Viewpoint), so it's not
always zero. In view3dscene/Castle Game Engine, I actually use the
bounding box center as the default pivot for Examine camera, so
Examine navigation works "out of the box" for any scene.

>From my experience (coding various 2D games using CGE :) ), it is
useful that a 2D orthographic projection is defined as 4 corners, not
just width,height. It allows to shift the center, which means that the
meaning of "OrthoViewpoint.position" is configurable.

E.g.

- with OrthoViewpoint.fieldOfView = (-1, -1, 1, 1), the
OrthoViewpoint.position determines what is visible precisely at the
center of the screen.

- with OrthoViewpoint.fieldOfView = (0, 0, 2, 2), the
OrthoViewpoint.position determines what is visible precisely at the
bottom-left corner of the screen, for a square window. For panoramic
window (window.width > window.height), the OrthoViewpoint.position at
least determines what is visible precisely at the bottom of the
screen.

All these options are useful, i.e. they allow to "think" about a 2D
camera easier, in my experience. In some games, I want to place things
around the center of the screen. In other games, I want to place
things around the bottom of the screen, e.g. because hero is standing
at the ground (https://store.steampowered.com/app/746750/The_Unholy_Society/
:) ).

I think that, technically, we could indeed get rid of this (and define
"OrthoViewpoint.fieldOfView" as only 2 floats, and add any necessary
shift to "OrthoViewpoint.position"), but it's just easier to think
when it's 4 values. It's may be also more natural to animate
"OrthoViewpoint.position" in some cases.

A simple OrthoViewpoint test for CGE is on
https://github.com/castle-engine/demo-models/blob/master/navigation/ortho_viewpoint.x3dv
.

Regards,
Michalis