well, the way you'll need to do it is different. I simply took your image above and shuffled blocks and lines in the picture around using paint by hand.
For you, It is somewhere in the block area // do any processing and the stucture of your cvCapture in your driver.
What I notice though is that you image is outputting the data in lines and blocks.
I assume there is a sequence # that is coming back from the cam with a block(seq, width, height) and then the data with the lines having a height of 1.
Ignoring that extra non-image data is what gives you your green background as the image is larger than picture pixels.
The above has to be a result of the driver you are using, either a non-initialization of some part of the capture, or you not "processing" something in your
// do any processing block of your code.