Capturing and reproducing real realistic objects in any virtual environment is complicated and time-consuming. Imagine that to simplify this task, use a conventional camera with a built-in flash – from any mobile or digital camera. The global team of computer scientists has developed a new method that replicates the physical objects of the virtual and expanded reality space using only the instantaneous flash camera, without the need for additional and often expensive hardware support.
"In order to truly reproduce the object of a real object in the VR / AR environment, we need to process the 3D geometry and appearance of the object," says Min H. Kim, Associate Professor of Computer Science at KAIST in South Korea and lead author of the study. "Traditionally, 3D artists have either done it manually, it's a laborious task, or using specialized expensive hardware. Our method is simple, cheaper and more efficient, and reproduces real 3D objects simply by shooting from one camera with a built-in flash."
Kim and his collaborators Diego Gutierrez, professor of computer science at the Universidad de Zaragoza in Spain, and KAIST doctoral students Giljoo Nam and Joo Ho Lee will present this new job at SIGGRAPH Asia 2018 in Tokyo from December 4th to December 7th. The annual conference is represented by the most respected technical and creative players in the field of computer graphics and interactive techniques, as well as demonstrating the leading role of science, art, games and animation in other sectors.
Existing approaches for the purchase of physical objects require specialist hardware settings to achieve the desired geometry and appearance of the modeling. These settings may include a 3D laser scanner or multiple cameras or light dome with more than one hundred light sources. In turn, this new technology requires only one camera to create high-quality outputs.
"Many traditional methods of using one camera can only capture the 3D geometry of the object, but not the complex reflection of real-world objects provided by the SVRFD," notes Kim. The SVBRDF, which represents the spatially bi-directional reflection distribution function, is the main element of the real world's shape and appearance. "Using only 3D geometry, you can not reproduce the realistic look of an object in the AR / VR environment. Our technique can capture high-quality 3D geometry as well as the look of its material so that objects can be made in virtually any kind of environment."
The group demonstrated its system using the built-in camera of the digital camera, the Nikon D7000 and the Android mobile phone in several examples in its article "Practical 3D Object Obtaining with SVTBRDF without Structured Flash Photography". The new algorithm, which does not need the geometry of the target object input, successfully captures the geometry and appearance of 3D objects using the base and the flash and reproduces consistent results. The work featured various sets of subjects that covered a wide range of geometries and materials, including metals, wood, plastics, ceramics, resin and paper, and consisted of complex shapes, such as detailed detailed Nefertiti mini statutes.
In the future, researchers are hoping to further simplify the capture process or extend the method, for example, incorporating dynamic geometry or larger scenes.
Source of the story:
Materials provided Computer Equipment Association. Note: Content can be edited for style and length.