Method And Device For Encoding And Decoding A Video Bitstream Using A Selected Motion Vector

What is this patent about?

’129 is related to the field of video coding, specifically techniques for improving the efficiency of motion vector prediction during video decoding and encoding. Modern video codecs use motion estimation to reduce temporal redundancy between frames, but this requires transmitting motion vectors, which can consume significant bandwidth. The patent addresses the problem of reducing the memory requirements associated with storing motion vectors for use in motion vector competition.

The underlying idea behind ’129 is to reduce the number of motion vector candidates considered during motion vector prediction by applying a flooring function to the coordinates used to select motion vectors from previously decoded frames. This flooring function effectively reduces the resolution at which motion vectors from previous frames are sampled, leading to a smaller set of candidate motion vectors and reduced memory usage.

The claims of ’129 focus on a method and device for processing a video bitstream. The method involves determining motion vectors from neighboring blocks in the current frame, and determining a motion vector from a previous frame. The location in the previous frame is determined by applying a flooring function (implemented via bit-shifting) to the coordinates of the current block. These motion vectors are then used to generate a motion vector prediction candidate list , from which a motion vector is selected for decoding or encoding the current block.

In practice, the decoder (or encoder) first identifies motion vectors from spatially neighboring blocks in the current frame. Then, to incorporate temporal information, it calculates a position in a previously decoded frame using the flooring function on the current block's coordinates. This flooring operation, implemented with bitwise right and left shifts, effectively snaps the sampling location to a coarser grid. The motion vector at this snapped location in the previous frame is then added to the candidate list. Finally, the decoder selects a motion vector from this list to predict the current block's motion, using an index transmitted in the bitstream.

This approach differs from prior art by strategically reducing the number of motion vector candidates from previous frames. Instead of storing and considering every motion vector from the previous frame, the flooring function ensures that only a subset of motion vectors, sampled at a lower resolution, are included in the candidate list. This reduces memory bandwidth and storage requirements without significantly impacting coding efficiency, as the most relevant motion information is still likely to be captured by the coarser sampling grid. The patent also suggests only selecting motion vectors from previous frames if the corresponding block was inter-coded, further refining the candidate list.