NVIDIA Introduces Groundbreaking Camera Technology With Chimera -- World's First Mobile Computational Photography Architecture
Forging a path for the next wave of advanced, consumer-focused mobile imaging experiences and applications, NVIDIA today announced new capabilities delivered by the NVIDIA Chimera Computational Photography Architecture.
Available in the NVIDIA Tegra 4 family of mobile processors, Chimera architecture offers a number of features never before available on mobile devices, including always-on high-dynamic range (HDR) photos and videos, HDR panoramic and persistent tap-to-track capabilities.
"We developed the Chimera Computational Photography Architecture to take mobile photography far beyond where it is today," said Brian Cabral, Vice President of Computation Imaging at NVIDIA. "We're enabling developers and users to not only use image processing to enhance their photos, but also use computer algorithms to create images no lens can capture alone."
Previous mobile device architectures have made it difficult to use the best tools for different parts of complex image processing. Chimera architecture removes those boundaries by providing the power to conduct nearly 100 billion mathematical operations per second to perform image processing, using computational techniques used in X-ray CT scanners, deep space telescopes and spy satellites.
First revealed at CES 2013, the architecture redefines mobile imaging with always-on HDR photos and videos. This allows camera users to instantly capture high-quality, HDR images similar to how the human eye sees the world in a vast array of locations and scenes, and under diverse lighting conditions.
Additional new features include HDR panoramic, which takes wide-angle, or "fish-eye," shots that normally require an expensive digital single-lens reflex camera. The Chimera architecture captures a scene while the camera moves from side to side, up and down or diagonally effectively "painting" a panorama in real time from many angles and in any order the user wants. In contrast, competing offerings must either be moved in one direction along a single horizontal plane, or require significant amounts of post processing taking up to 35 seconds to stitch together the panorama.