MUMBAI, India, Aug. 24 -- Intellectual Property India has published a patent application (202641097530 A) filed by S. Murugaveni; K. Harisudha; and G. Elavel Visuvanathan on August 12, 2026, for Network-Adaptive Gan Image Compression For Real-Time Bandwidth Optimization In Edge Devices.
Inventors include S. Murugaveni; K. Harisudha; and G. Elavel Visuvanathan.
The application for the patent was published on August 21, 2026, under issue no. 34/2026.
Abstract: ABSTRACT NETWORK-ADAPTIVE GAN IMAGE COMPRESSION FOR REAL-TIME BANDWIDTH OPTIMIZATION IN EDGE DEVICES A network- adaptive image compression system for an edge device with limited bandwidth is disclosed. An edge-side learned encoder is used to convert an input image to its scalable latent representation consisting of multiple information elements. A network-state monitor is used to measure communication parameters, such as available bandwidth, latency, packet loss, jitter, and congestion. The adaptive controller determines the latent information, quantization precision, and/or coding parameters to depend on the network state and a reconstruction-quality condition. The compressed representation is then sent to a receiving device which uses an adversarially trained generator to reconstruct the image. Subsequent compression operating points can be modified by using communication and reconstruction feedback. The architecture adapts the transmitted image information to the existing communication resources and at the same time ensures that a predefined minimum reconstruction quality is maintained and that the use of bandwidth, packet retransmission, buffering and the communication load of the edge devices are kept low.
Disclaimer: Curated by HT Syndication.
