Google is doing things again, 3 d and capture black technology based on the technology of light!

'lights' role in the film and television works, games and virtual environment is crucial, it is an important means of modelling, in the film plays convey information and express emotions, foil atmosphere, portraying characters and mental changes, and so on. It affects the formation of the tone and style of the show, and the tone form relationship of unity of opposites, and other modelling method combined with the performance of the film rhythm and melody.

many special effects in film and television works are needs to be done with the help of a green screen, when the technology is not high, often happen 'five MAO effects' of the tragedy.

want to copy the perfect photosynthetic efficiency is a difficult problem. With the evolution of computer vision technology, computer has been able to more 'natural' to restore the face shape, skin texture, but on condition that a simulated lights lack of realistic sense.

and Google recently announced a new technology based on light - — Relightables。 To be able to get a perfect it.

Relightables system working process can be divided into three parts: capture, reconstruction, and rendering.

the core of this system relies on a contains multiple points of view ( The initiative) Spherical stereo depth sensor lights of spherical devices equipped with 331 of the 90 programmable lights and high resolution of 12. 4 mp camera reconstruction.

capture cameras used in the human body contains 32 infrared ( IR) Cameras and 58 RGB camera. Infrared sensors to provide accurate and reliable 3 d data, RGB camera quality geometry normal maps and textures. The frequency of 60 hz to record the original video camera, the researchers based on spherical gradient alternate use of two different illumination conditions.

next, the researchers upload data to the public in the repository, the first phase is to generate each 'slots' the depth of the figure, the segmentation image and 3 d grid [ Kazhdan and Hoppe 2013] 。

they use a sequence alignment algorithm to deal with the reconstruction of the grid, as a result, a long sequence can share common triangulation ( 三角) 。 The researchers put forward a new method to solve the problem of the choice of key frames, transform it into a MRF reasoning problems to solve. Each unique triangulation is parameters into the ordinary 2 d texture space, the space can be Shared with all the triangulation of the frame to be Shared.

the end of each grid has two available image gradient spherical lighting, which can generate albedo, normals, glossiness and ambient light block diagram. These figure compatible with standard rendering engine, can be used in any set of lighting conditions to regenerate the rendered image.

Google this new system can perfect light and shadow, the reduction of the hero of making synthetic images look more realistic. Seamlessly will capture the fusion to the real world of the human body or movies, games, etc. The number in the scene. It may be possible to completely change the 3 d capture technology.