PS5, a kind of DLSS by Sony emerges from a patent

PS5 could take advantage of a real-time image reconstruction technique comparable perhaps to Nvidia's well-known Deep Learning Super Sampling (DLSS), at least based on what emerges from a patent registered by Sony but of which there are no official communications. >
The documentation relating to the patent is in Japanese, which does not make its interpretation simple, but the technology seems to concern an image processing technique similar to the DLSS. In practice, it would be a technique capable of acquiring information on images and reconstructing them through machine learning, applying additional features.

In fact, the system should start from a lower resolution image to reconstruct it in order to provide high resolution output up to 4K, just like DLSS does. It could simply be an evolution of the classic checkerboard rendering, but the patent is registered in 2019, therefore it seems to be a specific PS5 technique.

On Nvidia GPU, the DLSS exploits the work of the tensor cores, or hardware elements specifically designed to perform additional operations not present on PS5 and Xbox Series X, while on the SoC of PS5 such an operation could be based on shaders and therefore potentially be more expensive in the general economy of GPU and CPU exploitation, in any case it is a solution that can give very convincing results without requiring the effort to maintain a very high native resolution.

We therefore await further information on the matter, in case it is officially confirmed and explained by Sony.

Source