Image and Video
A New Generation Video and Image Encoder
Video and Image are the main expression of media contents, they take over 90% of transmit bandwidth in Internet. Video encoding technology help to compress video and image up to 1/200 of original size, it emits the temporal and spatial redundance through block based compression technology. H.264 and H.265 are two most popular video encoders. They published in 2003 and 2013 separately.
The 3 new generation Video and Image Encoder standards are AV1, VVC(H.266), AVS3, they can save 30% file size compare to H.265. We hope to introduce the new generation technology into Taobao short videos, Taobao live and Taobao pictures.
DL based Video and Image compress technology is another branch, it takes advantage of neural network to provide end to end compression capacity. We notice that DL based method gets higher BD-Rate gain in some use case.
3D point cloud compression is another technology. There are many useful cases in Taobao “TanPing” business. We hope to use this technology to compress 3D-model smaller.
- A real time H.266/AV1/AVS3 software encoder in personal computer, or hardware encoder with FPGA. 30% BD-rate gain compare to H.265;
- A real time H.266/AV1/AVS3 software decoder in X86 and Arm chipset. Especially in web browser.
- A DL based encoder and decoder, 10% BD-rate gain compare to H.265;
- More than 2 standard proposals imported by AV2 or H.267 or AVS3 enhancement extension.
- More than 1 CCFA paper is published.
Related Research Topics
- Content based video encode. Adaptive compress ratio according to video complexity and category.
- 6 DOF video encoding system, high efficiency of depth compress, viewpoint Rendering and hole filling method