You should not be modifying cameras.xml file by hand. The camera images are simply stitched width-wise by the PS3Eye CCV input component. This is done before CCV even sees the camera video data.
To stitch two cameras @ VGA you would set:
<WIDTH>1280</WIDTH>
<HEIGHT>480</HEIGHT>
<FRAMERATE>60</FRAMERATE>
To stitch two cameras @ QVGA you would set:
<WIDTH>640</WIDTH>
<HEIGHT>240</HEIGHT>
<FRAMERATE>60</FRAMERATE>
There is no frame doubling of any sorts. The images are simply stitched at the current frame rate.
The CCV consumes significant amount of CPU if not setup correctly, so play with the settings and use minimal amount of filters if possible.
Also use mini-mode when running a tracker. Image rendering in CCV is unnecessary once all is setup.
For more details, if you are a C/C++ programmer, you can look at the CCV source code.
Also if possible don’t use any gain settings. This introduces (amplify) sensor noise to the camera image.
AlexP