Google recently announced aDynIBaR Neural Dynamic Image-Based Rendering (NDIR) is a natural dynamic image rendering technology that automatically generates complex and dynamic scenes from a single video. It can also support bullet time, image stabilization, slow motion playback, and bokeh effects.
Compared to previous similar image rendering and reconstruction technologies, which must store complex scene data in a single structure and process it through a multi-layer perceptual neural network, once the video length and complexity increase, more computing resources will be required for processing, resulting in a decrease in computing efficiency and less than ideal actual processing results.
DynIBaR technology primarily records motion information and then uses artificial intelligence to automatically generate images from the user's desired perspective. This eliminates the need to generate entirely new images from scratch, instead leveraging existing pixel information to automatically generate new images. During processing, DynIBaR technology makes additional adjustments to optimize the image, ensuring consistent scene information across different time points. By rendering dynamic and static images separately, it improves overall image accuracy.
In terms of the application scope of this technology, Google said it can be used for handheld photography stabilization, which will be able to more effectively reduce image jitter and blur. It can also create smoother 3D image effects, or 5x slow-motion videos, and even present different bokeh effects, as well as image special effects such as bullet time.
Google said DynIBaR technology will be able to be used in mobile devices, allowing users to shoot clearer videos or create more video effects through mobile devices.
