Tile-Based VR Video Quality Estimation

The patent segments the VR video into multiple tiles with different quality levels (high-quality tiles in the viewing direction and low-quality tiles in other directions). The quality assessment is then performed separately for each tile type by extracting quality parameters from representative frames of each tile, and finally combining these segmented assessments to obtain the overall video quality. This segmentation approach enables accurate quality assessment that accounts for the different contributions of high-quality and low-quality tiles.

The patent applies local quality assessment by recognizing that different regions (tiles) of the VR video have different quality characteristics based on their viewing importance. High-quality tiles located in the user's viewing direction are assessed with higher weight, while low-quality tiles in peripheral directions are assessed with lower weight. This local quality differentiation allows the system to accurately reflect the actual user experience in tile-based distribution.

The patent implements local quality differentiation by distributing high-quality video data only for tiles in the user's viewing direction, while distributing low-quality or compressed data for tiles in peripheral directions. This approach ensures that bandwidth resources are concentrated on the most important regions that the user will actually view, thereby improving perceived user experience quality without proportionally increasing total distribution bandwidth.

The patent employs dynamic quality adaptation where the quality of distributed video tiles can change based on the user's viewing direction and head movement. As the user turns their head or changes viewing direction, the system dynamically identifies which tiles should be high-quality and which can be low-quality, adjusting the distribution strategy in real-time to optimize the balance between user experience quality and bandwidth consumption.