Samsung Electronics Co., Ltd.

Devendra Purbiya

Kiran Nanjunda Iyer

Ajit Shyamsunder Bopardikar

Hariprasad Kannan

Kausik Maiti

The present invention provides a method and system of generating multilayered point based 3D representation of a scene. The method includes constructing an alpha model of randomly selected regions associated with objects with similar depth value and depth jump in a depth map of a scene, and identifying mixed pixel regions associated with edges of the objects. The method further includes estimating color, depth and transparency values for pixels in each identified mixed pixel region using associated color information from an image map of the scene, associated depth value information from the depth map, and transparency information from the alpha model based on the associated depth information. The method includes generating a multi-layered point based 3D representation of the scene by organizing the estimated color, transparency, and depth values. Moreover, the method includes rendering 3D views of the scene using the multi-layered point based 3D representation of the scene.

Growing community of inventors

Yongin-si, South Korea

Devendra Purbiya