Nvidia introduces RTX VSR Video Super Resolution Technology: a Breakthrough in AI Image processing
CTOnews.com March 9 news, Nvidia recently launched the RTX VSR video super-resolution technology through the new driver. Today, the official article introduced the technology.
Officials say RTX VSR is a breakthrough in AI image processing, which goes beyond traditional edge detection and feature sharpening techniques to greatly improve the quality of live video content.
RTX VSR can reduce or eliminate artifacts caused by compressed video, such as blocking effect, ringing artifacts on edges, disappearance of high-frequency details and banding in flat areas, while reducing texture loss. It also sharpens edges and enriches details. This technique uses deep learning network to enlarge the image and reduce the compression artifacts in a single reasoning. The network analyzes low-resolution video frames and predicts residual images at the target resolution. Then, the residual image is superimposed on the traditional image enhancement to correct the artifact error and sharpen the edge to match the output resolution.
The deep learning network uses different degrees of compressed content for training. It learns the types of compression artifacts that do not exist in uncompressed images but exist in low-resolution or low-quality video, and uses these artifacts and their uncompressed images as a reference for network training. The Deep Learning Network also conducts extensive visual assessments to ensure that the generated models are suitable for almost all real-world and game content.
Currently, RTX VSR requires the use of GeForce RTX 40 or 30 series GPU for video content in Google Chrome and Microsoft Edge browsers. Chrome browsers (110.0.5481.105 or later) and Edge browsers (110.0.1587.56) have recently added RTX VSR support. To turn on this technology, open the NVIDIA control panel and open adjust Video Image Settings. Check the super-resolution box under "RTX Video Enhancement" and select one of the quality levels 1 to 4-the quality level represents the quality of the image enhancement and also depends on GPU performance.
