Optical Flow Processing for Efficient Video Data Transmission in Telecommunications

The paper presents a method for optimizing inter-frame prediction based on optical flow analysis and the selective activation of affine motion models. The proposed approach is designed to enhance the efficiency of video data transmission in telecommunication networks by reducing bitrate and latency while maintaining visual quality. The author addresses the practical challenge of reducing bitrate load and latency variability during uplink live video transmission in 5G/6G networks. A method referred to as curl-gated Flow is proposed, in which local affine motion models (both SIM(2) and the general six-parameter affine model) are selectively activated only in those frame tiles where the magnitude of the optical flow curl indicates a non-zero rotational or deformational component of motion.

Hali tarjima qilinmagan