siliconwhe.blogg.se

Now, you do the shift, and you find you're missing chunks of background that were covered in your source video, but not in your new shifted copy - exactly as if you'd cut the shapes out of paper, and shifted them over a bit - there's a hole. Once you have the z-depth for everything, you know how much you need to shift every object by to create the view from the 'other eye', by simple trigonometry. You might be able to do the above automatically, or semi-automatically, by analyzing the perspective cues, etc, but it's probably largely a manual process - matte out an object, propagate the matte over time, set the z-depth for that object - repeat and repeat. So what you do is decide, for every scene and every object on the screen, what depth it belongs to - pretend each object is a cardboard cutout of itself, and stack them up at the appropriate distances.

Well, keep in mind that the only difference in content between the two video streams in current 3D systems is:Ģ) The visibility of background elements covered by those objects.