360-Degree Video Encoding Boundary Artifact Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current 360-degree video processing techniques suffer from significant seam/boundary artifacts due to the transformation of three-dimensional surfaces onto two-dimensional maps, particularly when adjacent regions are not spatially adjacent, leading to rendering issues and quality degradation.
Innovation Solution
The implementation of guard bands and adjustments in video encoding, such as disabling in-loop filtering and modifying quantization parameters, to prevent seam/boundary artifacts by treating projection boundaries as non-adjacent regions during encoding and decoding processes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If in-loop filtering is applied across projection boundaries in 360-degree video encoding, then compression efficiency is improved, but seam artifacts and visual discontinuities are introduced at boundaries where adjacent regions are not spatially adjacent
Solution Approach 1:
The video picture is divided into multiple non-overlapping regions corresponding to different cube map faces, with projection boundaries identified between them. In-loop filtering is selectively applied within each region while disabled across projection boundaries, allowing compression efficiency to be maintained within regions while preventing seam artifacts at boundaries.
Solution Approach 2:
Different filtering strategies are applied to different spatial locations: in-loop filtering is enabled within regions where spatial adjacency exists to maintain compression efficiency, but disabled at projection boundaries where adjacent regions are not spatially adjacent to prevent seam artifacts. This local differentiation resolves the contradiction between compression efficiency and artifact prevention.
2Productivity
If intra-prediction and inter-prediction are applied across projection boundaries, then prediction accuracy and compression efficiency are improved, but discontinuities and artifacts are introduced because adjacent regions are not spatially adjacent
Solution Approach 1:
The picture is segmented into multiple regions with projection boundaries identified. Prediction operations (intra-prediction and inter-prediction) are permitted within regions but blocked across projection boundaries, ensuring prediction accuracy is maintained within spatially coherent regions while preventing discontinuities at boundaries where regions are not spatially adjacent.
Solution Approach 2:
Prediction accuracy is locally optimized within each region by enabling intra-prediction and inter-prediction, while boundary continuity is preserved by disabling prediction operations that would cross projection boundaries. This local quality differentiation resolves the contradiction between prediction accuracy and boundary continuity.
3Device complexity
If standard video encoding is applied without considering projection boundaries, then encoding simplicity is maintained, but seam artifacts and quality degradation occur at projection boundaries
Solution Approach 1:
Projection boundaries are identified and marked before the encoding process begins. This preliminary identification allows the encoder to pre-determine where to disable in-loop filtering and prediction operations, maintaining encoding simplicity by avoiding complex real-time boundary detection while still preventing seam artifacts at known projection boundaries.
Data Source
AI summary
Provided are systems and methods for processing 360-degree video data by obtaining at least one 360-degree rectangular formatted projected picture; detecting a projection boundary in the at least one 360-degree rectangular formatted projected picture; disabling at least one of an in-loop filtering, an intra-prediction, or an inter-prediction, based on detecting the at least one 360-degree rectangular formatted projected picture comprises the projection boundary; and generating an encoded video bitstream.