mue
04-21-2010, 03:08 PM
Hi!
In the recent past i played around with the idea to get 180 degree field of view with only one single projector. In the following i will describe my thoughts and the outcome.
Projection screen
For the screen there are basically two options:
1) curved screen
2) multiple (three) angled planar screens
I choose the second option, since planar screens are easier to build and the computing and configuration of the necessary image warping (more on this later) seems to be easier too. So I have the following setup:
3789
Every single screen covers a horizontal field of view (hfov) of 60 degrees, gives me a total fov = 180 degree. The vertical fov (vfov) depends on the height of the screens.
Views
To achieve the correct outside visualization you have to project three adequate views on this screens. Each view with hfov=60 degree and vfov=xx degree. One view in the direction 60 degree left on the left screen, one view straight on on the front screen and one view 60 degree right on the right screen. Note: the visualization is only correct for the viewer if he is located at the sweet-spot.
The three views have to be generated by the simulation software, in my case fs2004. In fs2004 you can display multiple views. The view directions can be configured in panel.cfg and the hfov and vfov can be set by resizing and zooming the view windows. Problem here is that fs2004 doesnt show the values of hfov and vfov.
Projection
For the type of projection there are again two options:
1)front projection
2)rear projection
In front projection the viewer sits between projector and screens. Therefore the projector has to placed high above the viewer. But this is not feasible for me because of room constraints. So I opted for rear projection:
3788
If you look from the projector position (rear projection) on the angled screens you will see this:
3791
And that's also the image the projector has to project on the angled screens to achieve correct display of the views. (With the difference that the image has to be mirrored because the projector works in rear projection mode.)
In the image above you see the side screens as a (very) smaller area than the front screen. That means fewer pixels can be projected on the side screens. This results in an inferior and darker picture on the side screens than on the front screen. A further problem is that the projector lose focus on the side screens. This compromise I'm willing to accept, since the viewer looks mainly to the front.
Image Warping
The three view windows the simulation software gives me have to be warped and arranged so that the projected views will match the projection screens. The warping and arrangement of the three views are specified by 8 control points (the corners of the screens).
3787
The user has to place the control points at the corners of the screens. The following picture shows the screens and the control points from another perspective.
3790
For copying, warping and arranging the view windows I wrote a proxy dll that intercepts directx calls. The dll further displays information of hfov and yfov of the view windows. (information I need for resizing and zooming the views to the right parameter to fit my projection screen setup). The proxy dll doesn't need much computing power: In "warping mode" my framerate only drops from 31 fps to 29 fps. Most work is done by the graphic card.
Here is a summery of the steps the user has to take:
-create the three views in fs2004
-resize and zoom the views to get the right hfov and vfov
-placing the 8 control points at the 8 corners of the screens
-activate the warping
I made a video where i try to show this concept. The image quality is bad despite HD resolution. It seems my cam doesn't like the light condition. The rear projection screens are only a test setup. They are made of 25 cm x 25 cm acrylglas with shower curtain on the projector facing side. Because of the acrylglas I get very bad reflections. But I hope you can see that the warping works. Note that the camera was not located at the sweet-spot so the views doesn't look correct (e.g. "horizon breaks" at the screen edges).
Here is the video:
http://www.youtube.com/watch?v=bqyhV2iJZhY
In the recent past i played around with the idea to get 180 degree field of view with only one single projector. In the following i will describe my thoughts and the outcome.
Projection screen
For the screen there are basically two options:
1) curved screen
2) multiple (three) angled planar screens
I choose the second option, since planar screens are easier to build and the computing and configuration of the necessary image warping (more on this later) seems to be easier too. So I have the following setup:
3789
Every single screen covers a horizontal field of view (hfov) of 60 degrees, gives me a total fov = 180 degree. The vertical fov (vfov) depends on the height of the screens.
Views
To achieve the correct outside visualization you have to project three adequate views on this screens. Each view with hfov=60 degree and vfov=xx degree. One view in the direction 60 degree left on the left screen, one view straight on on the front screen and one view 60 degree right on the right screen. Note: the visualization is only correct for the viewer if he is located at the sweet-spot.
The three views have to be generated by the simulation software, in my case fs2004. In fs2004 you can display multiple views. The view directions can be configured in panel.cfg and the hfov and vfov can be set by resizing and zooming the view windows. Problem here is that fs2004 doesnt show the values of hfov and vfov.
Projection
For the type of projection there are again two options:
1)front projection
2)rear projection
In front projection the viewer sits between projector and screens. Therefore the projector has to placed high above the viewer. But this is not feasible for me because of room constraints. So I opted for rear projection:
3788
If you look from the projector position (rear projection) on the angled screens you will see this:
3791
And that's also the image the projector has to project on the angled screens to achieve correct display of the views. (With the difference that the image has to be mirrored because the projector works in rear projection mode.)
In the image above you see the side screens as a (very) smaller area than the front screen. That means fewer pixels can be projected on the side screens. This results in an inferior and darker picture on the side screens than on the front screen. A further problem is that the projector lose focus on the side screens. This compromise I'm willing to accept, since the viewer looks mainly to the front.
Image Warping
The three view windows the simulation software gives me have to be warped and arranged so that the projected views will match the projection screens. The warping and arrangement of the three views are specified by 8 control points (the corners of the screens).
3787
The user has to place the control points at the corners of the screens. The following picture shows the screens and the control points from another perspective.
3790
For copying, warping and arranging the view windows I wrote a proxy dll that intercepts directx calls. The dll further displays information of hfov and yfov of the view windows. (information I need for resizing and zooming the views to the right parameter to fit my projection screen setup). The proxy dll doesn't need much computing power: In "warping mode" my framerate only drops from 31 fps to 29 fps. Most work is done by the graphic card.
Here is a summery of the steps the user has to take:
-create the three views in fs2004
-resize and zoom the views to get the right hfov and vfov
-placing the 8 control points at the 8 corners of the screens
-activate the warping
I made a video where i try to show this concept. The image quality is bad despite HD resolution. It seems my cam doesn't like the light condition. The rear projection screens are only a test setup. They are made of 25 cm x 25 cm acrylglas with shower curtain on the projector facing side. Because of the acrylglas I get very bad reflections. But I hope you can see that the warping works. Note that the camera was not located at the sweet-spot so the views doesn't look correct (e.g. "horizon breaks" at the screen edges).
Here is the video:
http://www.youtube.com/watch?v=bqyhV2iJZhY