A Sony Interactive Entertainment patent application published October 29th describes a way to send less image data during cloud gaming. The proposed system removes selected sections of a game frame before transmission. Another AI rebuilds those sections on the receiving device.
The application, titled Image Decimation and Reconstruction Techniques, was highlighted by Patentlyze and published as US 2026/0340791 A1. It uses one machine-learning model to remove image data and another to rebuild the picture.
AI Decides Which Parts of a Frame Can Be Removed
In Sony’s example, a scene containing people and trees has more important visual details than much of the surrounding background. An importance map helps the AI identify what to keep and what could be rebuilt later.
The first AI model receives the image and can use that map to choose where to remove information. The resulting gaps are called holes, and the filing refers to the removal process as decimation. Rather than sending the original frame intact, the system sends a version with less image data.
But it doesn’t leave the receiving device to guess everything. Small pieces of the original image, called anchor elements, can be placed inside the gaps. These might be single pixels or groups of pixels that help the second AI work out what belongs there.
Without those reference pixels, the receiving AI could add details that weren’t in the original frame. The anchors give the reconstruction model a reference. That could allow more of the frame to be removed without sacrificing as much accuracy.
Two AI Models Handle the Stream
The first model runs on the sending side and creates the reduced image. That image can be compressed using a conventional codec before it travels over the network. On the receiving device, the second model fills in the missing areas using the image data and anchor elements that remain.
Sony also describes training the two models together using video game frames. That would help them learn which details can be removed while still producing a reconstructed picture close to the original. Target image quality and compression settings could influence how much data is cut from each frame.
Sony specifically mentions video game streaming through a cloud computing system as one possible use. It also describes virtual reality and mixed reality devices as possible receivers. The patent isn’t restricted to one PlayStation product.
Cloud Gaming Gains Still Need to Be Proven
Sending less image data could help on connections with limited bandwidth. The patent also suggests the technique could reduce latency. The application includes no performance measurements showing how much bandwidth it saves or how quickly a device can rebuild each frame.
We covered another Sony approach in January: CAE-IGS research that uses encoder information to adjust streaming resolution. Sony reported no measurable added encoding delay in those tests. This patent takes a different approach by removing parts of frames before transmission and filling them in afterward.
Sony hasn’t said whether this technology will make its way into PlayStation cloud gaming. Until we see it running with actual games, it’s hard to say how much of a difference it’ll make.
As always, remember to follow us on our social media platforms (e.g., Threads, X (Twitter), Bluesky, YouTube, and Facebook) to stay up-to-date with the latest news. This website contains affiliate links. We may receive a commission when you click on these links and make a purchase, at no extra cost to you. We are an independent site, and the opinions expressed here are our own.
