Improved depth-map generation methods for user-assisted 2D to 3D image and video conversion

点击翻译按钮获取中文摘要

The quality and the experience of a 3D stereoscopic movie or image provides new value for which consumers are willing to pay a premium. Increasingly, more movies are being screened in 3D as thousands of theatres become equipped to display 3D, as well as the availability of 3D consumer televisions cameras and projectors. With this comes a demand for content and tools to create and edit that content. However, there is a plethora of 2D non-stereoscopic content available, of which there is a demand to convert to 3D. Movie studios see great opportunity in converting many existing 2D movies to 3D as this would increase their consumer market. Currently, conversion of 2D to 3D is extremely laborious and mostly done manually by trained compositing and editing experts. It takes months to convert minutes of footage. Furthermore, the conversion of content has to be accurate and of high quality, while also being feasible. This is becoming an increasing concern as consumer scrutiny of 3D increases as the format becomes increasingly accepted. The work proposed in this research will help to create algorithms and tools that will enable more efficient and accurate conversion of 3D stereoscopic content from existing 2D content, with minimal and intuitive user assitance. The outcome is anticipated to be much improved methods, both in accuracy and in simplicity of use, for depth map generation than what is currently available. Furthermore, the proposed research is anticipated to eventually enable content creators or consumers to fine-tune the 3D experience by selectively adjusting depth for entire scenes or individual scene objects via simple controls.