Picture Resampling Filter Metadata for Heterogeneous Video
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video compression standards, such as VVC, lack effective mechanisms for specifying resampling filters, particularly for video sequences with heterogeneous resolutions, leading to inadequate homogenization of picture resolutions during post-filtering processes.
Innovation Solution
The method involves encoding metadata within a bitstream to specify resampling filters, including separable filters for luma and chroma components, and applying these filters only to pictures with resolutions different from the maximum resolution, using various filtering methods like bilinear, Directional Cubic Convolution Interpolation, and deep learning-based techniques.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If post-filtering SEI messages are used to improve picture quality, then subjective quality is improved, but they cannot specify resampling filters for heterogeneous resolutions
Solution Approach 1:
The patent applies different filtering parameters to different pictures based on their resolution characteristics. The filter parameters are adapted locally for each picture rather than applying a uniform filter to all pictures, enabling resampling of heterogeneous resolutions while maintaining quality.
Solution Approach 2:
The patent introduces dynamic filter parameters that can be adjusted according to the specific resolution and characteristics of each picture. The filtering process becomes adaptive rather than static, allowing the system to handle varying resolutions effectively.
2Adaptability or versatility
If resampling is applied at block level during decoding, then temporal prediction between different resolutions is enabled, but no resampled picture is available at decoder output
Solution Approach 1:
The patent performs resampling as a preliminary action during the decoding process rather than as a final post-processing step. The resampled pictures are prepared in advance and made available at the decoder output, enabling both temporal prediction and direct output usage.
3Ease of manufacture
If uniform filtering is applied to all pictures, then processing simplicity is maintained, but pictures of different resolutions cannot be homogenized
Solution Approach 1:
The patent implements local adaptation of filtering parameters for each picture based on its resolution. Instead of uniform filtering, each picture receives customized filter parameters that match its specific characteristics, enabling effective resolution homogenization.
Solution Approach 2:
The patent changes the filtering parameters dynamically according to the resolution of each picture. By adjusting parameters such as filter kernel size and coefficients based on picture dimensions, the system achieves resolution homogenization while maintaining processing efficiency.
Data Source
AI summary
A method comprising: decoding (701) a picture of a plurality of pictures representing a video sequence from video data; obtaining (702) parameters of a filter determined from metadata associated to the video data, the metadata comprising at least one first information specifying a subset of the plurality of pictures on which the filter is to be applied; and, applying (703) the filter on the decoded picture responsive to the metadata.