The patent of Jim Jannard, Jarred Land and Bhat Manjunath Subray was last updated in January 2019 and deals, among other things, with how RAW compression of moving images on a smartphone can be managed as resource-efficiently as possible.
If we understand it correctly, it is suggested NOT to calculate motion vectors in hardware (like H.264), but to use a discrete cosine transformation. The latter is quite computation-intensive, which is why early video doecs such as MPEG1 and MiniDV already rely on DCT. Even JPGs use the same principle. According to our memory, RAW compression of still images has also existed for a long time. We think we once read that already the first Canon Compressed RAW in the cameras is based on DCT. For this, only components other than RGB or YUV are fed into the DCT. And also the compression of CinemaDNG is, to our knowledge, based on DCT.
One news is that the hydrogen will internally record a compressed RAW dialect. And the other is that it will probably not be R3D, because it is based on wavelet compression, which is almost diametrically opposed to DCT.