Thursday, October 28, 2010

2D-to-3D Conversion: Identifying the Good and the Bad

SMPTE put the 2D-to-3D conversion process under a microscope Thursday, during a session that explored how to identify good and bad 3D, and examined various conversion techniques as well as how they can be used to enhance the storytelling process.
“The use of 2D-to-3D conversion has become an important if not controversial process,” said session chair Brad Hunt of Digital Media Directions. “It has become an integral tool in creating 3D content.”
Lack of quality of 2D-to 3D conversion is often caused by budgetary constraints, believes Matthew DeJohn of In-Three. Yet he suggested that 3D can vary from highly realistic to extremely unsettling.
DeJohn outlined three stages for conversion: Roto/image segmentation, depth creation, and the paint phase.
He discussed and demonstrated some approaches that many lend themselves to lower quality conversion:
--Rubber-sheet approach: Tends to avoid roto by creating a single piece of geometry.
--Watery or messy edges: Artifacts due to an automated fill. This is often used to cut costs in the paint process.
--Cardboard cut out approach: Used to cut costs on generating accurate models.
--Other issues: Lack of transparency, miniaturization and a “composited” look.
Asked what are the opportunities to use automated tools that might reduce cost, DeJohn suggested technology for the paint stage. “There are algorithms out there,” he said. “But they are not that advanced at this point.”

Stereo D’s head of stereography Graham Clark emphasized that 2D-to-3D conversion is artistic. “We treat it as a creative process and there is almost no automation.”
For conversion, Clark reported that one should consider: The lens, story, context, sound, depth cues, perspective, occulusion, atmospheric/aerial perspective, lighting and shade, texture/texture gradient, and motion parallax.
As an example of Stereo D's work, the “Thor” trailer was screened.

No comments:

Post a Comment