Author |
Message |
David Sykes
Rating: N/A Votes: 0 (Vote!) | Posted on Sunday, April 03, 2005 - 10:44 pm: | |
Assuming that you can use quad-buffered OpenGL, that is a fairly easy way to do stereo-compositing. Using shutter-glasses, you can adjust the stereo-depth of the objects you are compositing in 'realtime'. Simple methods are restricted by the size of the texture memory, maybe only 2048x2048. Are there any ways to work with images larger than the size of the texture memory and then save them ? |
M.H.
Rating: N/A Votes: 0 (Vote!) | Posted on Monday, April 04, 2005 - 4:46 am: | |
Yes. 1) Do not use textures. Use only normal buffers. 2) Use beter graphic card supporting biger textures 3) Use multiple textures for one image 4) Une NPR (non power of 2) textures having diferent limits 5) Use OpenGL only for preview. Do final rednering for saving by non-OpenGL methods. I can give you a ver very long list of other techniques. Just give more informations about what do you want exacly to do .... |
David Sykes
Rating: N/A Votes: 0 (Vote!) | Posted on Tuesday, April 05, 2005 - 1:43 am: | |
I have a stereo background image. (If the background is a long way away and has no depth I will use the same image for left/right views). I have objects I wish to stereo-composite. These images are TIFF or PNG files with transparency. They are rendered onto the background left/right quads, reduced in size if necessary and then postioned in stereo-space by altering the separation of the quads. Then the background and overlay should be merged into a single image. Add the next object to new background, resize and position, merge into single image. Repeat for all objects to be composited. Save as left/right images of chosen size (which may be larger than the texture-size). |
M.H.
Rating: N/A Votes: 0 (Vote!) | Posted on Tuesday, April 05, 2005 - 5:32 am: | |
David: You should study how to do OpenGL rendering direcly in a BMP, not to the screen. This gives you no limit for output image size ... There probably exist some other sample codes, but you can extract the neccesary from my work as well (follow the save as Bmp menu option): http://www.ccp14.ac.uk/ccp/web-mirrors/marchingcube-fourierviewer/~husakm/Public/MarchingCubeELD/MCE_MFC.HTM |
M.H.
Rating: N/A Votes: 0 (Vote!) | Posted on Tuesday, April 05, 2005 - 7:12 am: | |
In other words you should do following: - use standard OpenGL HW accelerated mode for preview - do not use textures for the big background images - use simple pixel transfer to the buffers - use hi-res textures for the object in foreground When you will be ready with the alighment switch to off screen rendering to hi-res BMP object as mentioned in the post above. OpenGl page flipping , nor L/R buffering will not work - just simpy render 2x to 2 BMP with identical camera parameters as used for the HW accelerated preview ... What you speek about is a big project requesting several months of work to fine tone everything. Good luck |
M.H.
Rating: N/A Votes: 0 (Vote!) | Posted on Tuesday, April 05, 2005 - 7:14 am: | |
Another technique you can catch from MCE_MFC code ... The final rendering to still images is done in POV-ray - by exporting the OpenGL scene geometry into an Povray file ... This is another method how to get hi-quality final renderings ... |
David Sykes
Rating: N/A Votes: 0 (Vote!) | Posted on Tuesday, April 05, 2005 - 10:24 am: | |
Thanks Michal. I will study the code. It would be FAR easier if I could persuade Masuji to do it :-) For some reason, it would require a rewrite of SPM. That is far too much work. David |
M.H.
Rating: N/A Votes: 0 (Vote!) | Posted on Tuesday, April 05, 2005 - 12:22 pm: | |
What you probably want to do is a 2D-3D conversion. Exacly what you need is already fully implemented in the form of Adobe Photoshop plugin - part of Syntha Gram Content Creation kit avaliable from Stereographics. They use Direct3D for the work but it works identicaly as unde OpenGL. The code is distributed with SynthaGram monitors ... So if you know anybody with SynthaGram monitor avaialbe, ask him for a demo. You will save your work by not-duplicating already fully fine toned code development. |
Anonymous
Rating: N/A Votes: 0 (Vote!) | Posted on Sunday, June 05, 2005 - 5:06 am: | |
hello, I am trying to render a texture mapped Quad in stereo. I can draw the Quad and see it correctly in stereo but when I apply the texture to the quad it does not work. WHen I run the application it flickers. and random patches of color are shown. I do not know why is the problem ? I am using anaglyph for stereo rendering. Please reply to tanphadke@yahoo.com THnaks |
M.H.
Rating: N/A Votes: 0 (Vote!) | Posted on Sunday, June 05, 2005 - 8:29 pm: | |
thanspadke: 1) Put in your code standard OpenGL error diagnostic 2) If you will not find the problem in this way post the code (in the most simpyfied version still showing the probem) 3) Try the discusin board on www.opengl.org If this is a comercial problem, my company can offer development+debuging on comercial base. |
David Sykes
Rating: N/A Votes: 0 (Vote!) | Posted on Sunday, June 05, 2005 - 11:31 pm: | |
Which OpenGL command are you using ? Some commands require a power-of-two texture-dimension. Incidentally, all replies to a post should be to this forum so that everyone can benefit from the replies. |
David Sykes
Rating: N/A Votes: 0 (Vote!) | Posted on Monday, June 20, 2005 - 10:22 am: | |
> The final rendering to still images is done in > POV-ray - by exporting the OpenGL scene geometry > into an Povray file I have studied your save BMP and save PovRay in MCE_MFC code. PovRay seems a nice way to do it. I have rendered the background onto a plane and overlaid with a TIFF image. TIFF transparency is automatically supported. I will now decide how to record the position of each overlay image and generate the geometry file. David |
Anonymous
Rating: N/A Votes: 0 (Vote!) | Posted on Monday, June 20, 2005 - 9:26 pm: | |
Yes. I have 2 methos: OpenGL rendering to BMP (instead of to screen) Export to POVRAY input files ... |