2D3D Image Conversion

Both the left viewpoint image and right viewpoint image are required for three-dimensional (3D) TVs. Toshiba has developed a 2D3D image conversion technology that creates the two viewpoint images from one input image.

Depth structures of the scenes are generated using the following three image recognition technologies:

Background estimation

The input image is classified into one of a number of typical scenes and its corresponding depth data are applied.

Motion detection

A faster object is placed at a nearer depth.

Human detection

A human 3D model is applied to the detected positions of faces.

Next, each viewpoint image is calculated using the depth structure. 2D3D image conversion technology makes it possible to view standard 2D contents in 3D, in addition to 3D contents themselves, and is being incorporated into TVs and PCs in the Japanese market.

2D3D image conversion technology that creates nine viewpoint images from a single or two viewpoint images is also being incorporated into our glasses-free 3D TV models, which are the world’s first 3D TVs^(*) that do not require dedicated glasses.

(*) As of October 2010 (as researched by Toshiba)

Flow of 2D3D conversion