A discussion regarding Softimage 2011.5 some weeks ago prompted me to write this post. I inquired as to why Softimage has standardised to only offer side-by-side stereo pairs (be it cross or parallel) instead of the Over-Under format, their answer was simple,
‘To be honest we spoke to a few customers about their stereo requirements and you are the first (that I am aware of) to mention over-under for stereo previews. Could you maybe explain the advantages of over-under for you guys and why you settled on that standard?’
Now I truly feel that the over-under method for storing stereo paired video footage makes far more sense than side-by-side… and here are my reasons.
Often with stereography you require HIT (Horizontal Image Translation) in post, effectively sliding your images left or right to move the depth bracket back and forth in Z Space. Now with images stored in a side-by-side format at 1920×1080, each eye is halved (960×1080) on the horizontal resolution this to start with is a bad as our eyes are separated horizontally so any loss in pixel resolution on the horizontal has the most direct effect to viewing quality when re sampling these images back to 1920×1080.
Secondly and referring to the HIT if you slide your left image right & right image left your pixels are crushed down the center and lost because the stereo re-sampling method always cuts at the halfway mark of your image resolution. Over-Under avoids both of these issues by maintaining full Horizontal Pixel Resolution at 1920×540, and also avoids destructive behaviour to the resolution of the image because the HIT happens down the Horizontal not the Vertical. (sounding very outer limits there, lol)
The introduction of black pixels down the edges of the frame occur in both side-by-side and over-under formats, however they are overcome by the over rendering of frames. We render a buffer of 20px on either side allowing for an overall HIT of 40px left & right. The most important fact to note here is the destructive tendencies of side-by-side footage when images are HIT inwards, and the fact that most editing software assumes that anything beyond one side of the halfway mark on the horizontal is the opposite eye therefore even with the over rendering of pixels, black bars centrally are unavoidable.
Thirdly, many of the new passive monitoring systems such as the Zalman Trimon 3D Display and the JVC GD-463D10 use a horizontal interleaving polarization method. This effectively makes every alternate row of pixels polarized for a specific eye. When converting a side-by-side stereo pair to be viewed on a interleaved monitor you loose a quarter of your total video resolution. The side-by-side video is halved on the horizontal (1 half lost) and then when the monitors interlace the footage each alternate row of pixels is lost due to it being used to show the opposite eye (1 half lost) this effectively means you only see a quarter of your total resolution.
The Over-Under format stores the images with full horizontal resolution and has the pre-compensated for the loss of pixel resolution that occurs during the interleaving process. Therefore there is no loss of pixels. We use Stereoscopic Player to preview the captures in stereo weather it be on our Polarised Horizontally Interleaved Monitors or Active Projector (Quad Buffered) as it deals with converting the input formats to the required output methods. The same applies in our edit suit where we monitor stereoscopic on a Panasonic TH-P50VT20P
Just on a final note, the HDMI 1.4 stereoscopic standard is two full resolution frames Over-Under…