Video signals may be noisy. Noise may reduce the temporal correlation between frames, which in turn reduces compression efficiency in codecs. To increase the temporal correlation between frames of the video signal, temporal filtering may be performed. Novel temporal filtering techniques for real-time applications are described in this document. In real-time applications, encoding speed is critical; and memory usage may be a consideration. Two main aspects are described; namely, an in-place temporal filtering and an in-loop temporal filtering algorithm. In-loop filtering includes that temporally filtered frames are themself (instead of source frames) used for temporally filtering subsequent frames. Encoding statistics can be used to guide the temporal filtering process. The encoding statistics include, for example, quantizers of current frames and motion vectors of blocks of previously coded filtered frames. In-place temporal filtering means that the source frames are themselves modified by the temporal filtering process. This results in time and memory savings.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Wang, Yunqing, "An In-Loop & In-Place Temporal Filtering For Real-Time Applications", Technical Disclosure Commons, (March 09, 2022)