By Steve Shaw, Light Illusion
Last time in Part2 of this series, we looked at convergence and the pros and cons of parallel versus converged shooting. I concluded that, all things considered, I prefer the parallel approach in order to avoid keystoning problems and time lost during shooting. This month I will discuss issues concerning the edges of the screen, including edge violations and floating windows.
When viewing stereoscopic images on a small-sized screen, i.e. smaller than IMAX, the issue of the edges of the screen is something that must be considered, especially left and right of screen.
The problem with the above image is that the camera dolly is exiting the screen frame to the right and bottom, while positioned in front of the screen plane (i.e. it has negative parallax). This destroys the illusion of 3D, because your brain tells you that any object in front of the screen plane should not be cut off by the screen edges. In fact it is the left and right of screen that causes the real problem, rather than the top and bottom.
If the object exits the screen quickly, there probably is not too much of a problem since the brain won’t have time to register the issue, but if the object lingers or moves slowly out of frame it will cause a problem.
In reality, there are usually few times that objects should be allowed to extend beyond the screen plane unless the object is fully contained within screen frame surround, and even then, such ‘in your face’ 3D effects can easily detract from the story. So use this effect with caution!
Another major issue with screen edges is when an object placed in front of the screen plane (i.e. it has negative parallax) and exits the screen in one eye’s image before the other. In this instance, the illusion of 3D is broken, since without both the left and right eye containing the same portion of the image, there can be no 3D effect.
The only real way to overcome this issue is to either change the convergence point, such that all objects are behind the screen window (i.e. have positive parallax), or to manage it in post by ‘zooming’ into the centre of the image in order to remove the problem object at the screen boundary from both the left and right eyes.
However, a more dynamic way to deal with the problem is to dynamically crop just the eye that has more of the object such that it matches the other eye.
This approach has become known as ‘floating windows’, or more accurately ‘floating crops’, as the edges of the left and right eyes are cropped dynamically depending on the objects at the edge of screen.
Note that although the chap in centre frame is well in from of the screen plane there is no issue as the top and bottom crop points are identical. If the camera were panning vertically however, there would be an issue as the crop points would no longer be identical.
When viewing stereoscopic images on a small-sized screen, edge violations in negative parallax are something that must be considered, especially to the left and the right of the screen, as they will cause a conflict in the brain between the depth of the object and the screen’s edge. This can be dealt with in post, but can be avoided by using negative parallax effects sparingly, and using it within the centre of the screen.
The floating window can be a useful way to manage issues with objects leaving the screen unevenly and causing stereoscopic failure. Indeed it is argued that a likely reason that objects leaving the screen unevenly causes problems has a lot to do with the evolution of human vision, and how historically predators were often glimpsed in the periphery of vision. A disturbance to the left or right therefore brings back natural survival instincts and causes stress. This might also explain why the same is not true with upper or lower periphery vision – there were not so many predators there!
Next month I will discuss interaxial distance, miniaturisation and convergence and focus.
Steve Shaw is a Partner in Light Illusion, a top consulting service for the digital film market, with offices in the UK and India.