Adaptive Transform Kernel Sizes for Video Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding technologies face challenges in efficiently compressing video data while maintaining video quality, particularly due to limitations in bandwidth and memory resources, and there is a need for improved methods to adaptively determine transform parameters for better hardware implementation.
Innovation Solution
Adaptive determination of transform parameters, such as kernel sizes and shapes, for both primary and secondary transforms, using coded information known to both encoder and decoder, allowing for smaller sizes when appropriate, which enhances hardware efficiency and throughput.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If fixed transform kernel sizes are used in video coding, then hardware implementation is simplified, but video compression efficiency and adaptability to different content characteristics deteriorate
Solution Approach 1:
The patent applies dynamics by making the transform kernel size adaptive rather than fixed. The kernel size is dynamically determined based on coded information such as block size, prediction mode, and transform type, allowing the system to optimize compression efficiency for different video content characteristics while maintaining hardware implementability through standardized adaptation rules.
Solution Approach 2:
The patent changes the parameter of transform kernel size from a fixed value to a variable parameter that depends on coded information. By establishing relationships between kernel size and parameters like block size, prediction mode, and transform type, the system achieves flexible adaptation to different coding scenarios without requiring completely different hardware for each configuration.
2Productivity
If larger transform kernel sizes are used, then video compression efficiency improves, but hardware resource requirements and complexity increase
Solution Approach 1:
The patent applies local quality by selecting different transform kernel sizes appropriate for different local coding scenarios. Instead of using a universally large kernel that would maximize compression everywhere, the system chooses kernel sizes tailored to specific block sizes, prediction modes, and transform types, optimizing the balance between compression efficiency and hardware complexity for each local context.
Solution Approach 2:
The patent applies partial action by using transform kernel sizes that are sufficiently large to achieve good compression for most cases, but not excessively large to overwhelm hardware resources. The kernel size is carefully controlled to be appropriate for the given block size and transform type, avoiding unnecessary computational complexity while maintaining adequate compression performance.
3Productivity
If adaptive transform parameters are implemented, then video compression efficiency and hardware efficiency improve, but decoding complexity and processing time increase
Solution Approach 1:
The patent applies preliminary action by pre-establishing the relationships and rules for determining transform kernel sizes based on coded information. The adaptation logic is prepared in advance through standardized relationships between kernel size and parameters like block size and prediction mode, allowing fast lookup and computation during decoding without requiring complex real-time analysis.
Data Source
AI summary
An example method of video decoding includes receiving a video bitstream comprising a set of encoded blocks and a corresponding set of transform coefficients. The method also includes selecting a non-separable secondary transform from a set of one or more non-separable secondary transforms based an index value. The method further includes determining, based on coded information, a kernel size and kernel shape for the non-separable secondary transform, and decoding the set of encoded blocks by applying the non-separable secondary transform with the determined kernel size and kernel shape to the set of encoded blocks.