Scaling Window Range Expansion for Multi-Layer Video Resampling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing VVC text has restrictive value ranges for scaling window offsets and inadequate signaling controls for reference picture resampling (RPR), which hinder efficient decoding and encoding of multi-layer video streams, particularly in viewport-dependent 360° video coding schemes.
Innovation Solution
The scaling window offset value ranges are expanded to allow widths and heights greater than or equal to twice the picture dimensions, and signaling controls are refined to depend on layer relationships and specific parameter differences, enabling more flexible RPR operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If the scaling window offset value ranges are restricted to traditional limits, then the decoding process remains simple and computationally efficient, but the ability to perform reference picture resampling (RPR) for multi-layer video streams and viewport-dependent 360° video coding is hindered
Solution Approach 1:
The patent changes the parameter ranges for scaling window offsets by introducing new syntax elements (sps_scaling_win_left_offset, sps_scaling_win_right_offset, sps_scaling_win_top_offset, sps_scaling_win_bottom_offset) with expanded value ranges. These parameters allow the scaling window to extend beyond the traditional picture boundaries, enabling RPR operations where the scaling window width and height can be greater than or equal to twice the picture dimensions. This parameter expansion directly enables multi-layer video stream processing and viewport-dependent 360° video coding while maintaining a systematic approach to managing the increased complexity through standardized syntax element definitions.
2Productivity
If the scaling window width and height are limited to traditional ranges, then the bitstream remains compact and transmission efficiency is maintained, but the support for advanced decoding tools like DMVR and BDOF in multi-layer video is inadequate
Solution Approach 1:
The patent extends the scaling window into additional spatial dimensions by allowing offsets that position the window beyond the current picture boundaries. The scaling window is defined by four offset parameters (left, right, top, bottom) that can create a window larger than the picture itself, effectively adding dimensional flexibility. This enables reference pictures to be resampled from regions that extend beyond the visible picture frame, preserving more reference information for advanced decoding tools like DMVR and BDOF while maintaining efficient bitstream representation through systematic syntax element encoding.
3Measurement precision
If signaling controls for RPR are made inadequate or simplified, then the encoding process remains fast and computationally light, but the precision and control over reference picture resampling in multi-layer video streams is insufficient
Solution Approach 1:
The patent segments the RPR control into multiple independent syntax elements, each governing a specific aspect of the scaling window configuration. The four offset parameters (sps_scaling_win_left_offset, sps_scaling_win_right_offset, sps_scaling_win_top_offset, sps_scaling_win_bottom_offset) divide the overall scaling window positioning control into four separate, manageable components. This segmentation allows precise control over each dimension of the scaling window independently, enabling fine-grained RPR control for multi-layer video streams while keeping each individual syntax element simple and efficient to encode and decode.
Data Source
AI summary
Video processing, including video coding, video decoding and video transcoding, is described. One example method includes performing a conversion between a video comprising a video picture and a bitstream of a video according to a rule, where the rule specifies that syntax elements indicate a first width and a first height of a scaling window for the video picture, and where the rule specifies that a range of permissible values for the syntax elements include values greater than or equal to two times a second width of the video picture and two times a second height of the video picture.