Raw Video Formats
The website http://www.fourcc.org/ has more information than you ever need to know about all kinds of raw video formats. Only a few are relevant, such as the ones listed below. Raw video pictures in GStreamer are almost always in the following formats:
I420
Planar YUV format, with width x height bytes of Y component values followed by (width/2)x(height/2) bytes of U component values, followed by (width/2)x(height/2) bytes of V component values. The U and V components are scaled down by a factor of two in each direction.
YUY2
Packed YUV format, with (width*2) x height bytes of component values. Each line is a series of component values YUYVYUYV..., with the U and V components downsampled horizontally by a factor of 2.
AYUV
A packed YUV format with (width*4) x height bytes of component values. Each line is a series of component values AYUVAYUV... The component 'A' is for alpha values.
32-bit RGB and ARGB
A packed RGB format with (width*4) x height bytes of component values. Each line is a series of component values ARGBARGB...
YV12
Same as I420, but with U and V swapped.
YVYU
Same as YUY2, but with U and V swapped.
Stride
In most cases, the total number of bytes in a scan line is rounded up to a multiple of 4.
References
The canonical description of raw video formats in GStreamer is the source code for videotestsrc.
Stupid Weird Stuff
Interlaced I420
Lines 0, 2, 4... of the Y component belong to the upper field, and lines 1, 3, 5... belong to the lower field (obviously). However, this is also true for the U and V components, which is somewhat counterintuitive, since the U and V components are downsampled vertically. Thus, when reconstructing full-size U and V component images, you need to upsample each field individually, then reinterlace.

