Extended Motion Vector Predictor Set for HEVC Bitrate Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In video coding, the transmission of motion vectors requires a significant portion of the overall bandwidth, particularly in low-bitrate applications or systems with smaller blocks or higher motion accuracy, necessitating improved motion vector prediction techniques to enhance compression efficiency.
Innovation Solution
An extended motion vector predictor set is introduced, incorporating spatial and temporal candidates from neighboring and co-located blocks, allowing for more comprehensive prediction without additional side information transmission, and enabling implicit derivation at the decoder.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of energy
If motion vector prediction is used to reduce transmission bandwidth, then bitrate is reduced, but the complexity of prediction scheme increases
Solution Approach 1:
The motion vector prediction is segmented into two components: a predictor value derived from spatial and temporal neighbors, and a residual value that captures the difference. This segmentation allows the predictor to handle the majority of the motion information while the residual handles only the differences, reducing overall bitrate complexity
Solution Approach 2:
The prediction approach adds temporal dimension to the traditional spatial prediction by incorporating motion vectors from reference frames. This temporal dimension allows the system to exploit motion coherence across frames, improving prediction accuracy without proportionally increasing complexity
2Measurement precision
If more motion vector predictors are used to improve prediction accuracy, then coding efficiency is improved, but the amount of side information to be transmitted increases
Solution Approach 1:
Instead of transmitting all available motion vector candidates, the system transmits only the necessary residual information after prediction. This partial action approach provides sufficient prediction accuracy by transmitting only what is needed to correct the predictor, rather than all possible predictor information
Solution Approach 2:
The predictor is constructed by copying motion vector information from spatially and temporally neighboring blocks. This copying approach reuses existing decoded motion information without requiring additional side information transmission, as the neighbor block motion vectors are already available at the decoder
3Productivity
If temporal prediction is prioritized over spatial prediction, then coding gain is improved, but the complexity of predictor selection increases
Solution Approach 1:
The system merges spatial and temporal prediction approaches into a unified predictor construction process. Motion vectors from both spatial neighbors and temporal references are combined to form the predictor, allowing the decoder to benefit from both spatial and temporal coherence without requiring complex separate processing paths
Solution Approach 2:
The predictor is constructed in advance using readily available spatial and temporal motion information before the actual motion vector needs to be encoded. This preliminary construction of the predictor simplifies the encoding process and reduces real-time complexity by pre-computing the prediction baseline
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
A method and apparatus for determining a motion vector predictor (MVP) or a MVP candidate based on a MVP set are disclosed. In video coding systems, the spatial and temporal redundancy is exploited using spatial and temporal prediction to reduce the video data to be transmitted or stored. Motion vector prediction has been used to further conserve the bitrate associated with motion vector coding. Motion vector prediction technique being developed for the current high efficiency video coding (HEVC) only uses a MVP candidate set including spatial MVP candidates and a temporal candidate corresponding to the co-located block. In the current disclosure, the spatial and temporal motion vector predictor set is extended to include at least one spatially neighboring block associated with list 0 reference pictures and list 1 reference pictures, and a co-located block selected from all co-located blocks inside the co-located Prediction Unit (PU) except the one located at the top-left of the PU.