Graph-Based Transform Template for Video Coding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video coding methods rely on single linear transforms like DCT or DWT, which are inefficient due to high bit cost and bitrate overhead for transmitting transform coefficients, especially when dealing with next-generation high-resolution video content.
Innovation Solution
A method using graph-based transforms, where a template graph set is encoded for each frame, and a template index is transmitted to indicate the optimal transform for each coded block, allowing for adaptive and efficient coding.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple general linear transforms are used to fit specific statistical characteristics of signals, then video compression performance is improved, but bit cost and bitrate overhead for transmitting transform coefficients increase significantly
Solution Approach 1:
The patent uses a template graph that represents a simplified copy or model of the actual pixel relationships in a video block. Instead of transmitting full transform coefficients for multiple transforms, the system transmits a template index that points to a pre-defined graph template, which captures the essential statistical characteristics needed for transformation. This copying approach reduces bit cost while preserving compression performance.
Solution Approach 2:
The patent changes the parameter representation from full transform coefficients to a compact template index. By parameterizing the transform characteristics through graph templates with adjustable weights and structures, the system can adapt to different statistical characteristics while transmitting only the template index and modified parameters, significantly reducing bitrate overhead compared to transmitting complete transform data.
2Device complexity
If a single transformation (DCT, DWT, or Hartmard) is used, then computational complexity is kept low, but video compression efficiency is insufficient for next-generation high-resolution content
Solution Approach 1:
The patent introduces dynamic adaptability through graph templates that can be selected and configured based on the statistical characteristics of each video block. The template weights and structures can be adjusted dynamically to match local signal properties, enabling the system to adapt from simple to complex transforms as needed, thus improving compression efficiency without permanently increasing computational complexity.
Solution Approach 2:
The patent segments the video content into blocks and applies different graph template transformations to different segments based on their statistical characteristics. This segmentation allows the system to use simple transforms for uniform regions and more complex adaptive transforms for regions with specific patterns, optimizing the balance between computational complexity and compression efficiency across the entire video sequence.
Data Source
AI summary
A method for decoding a video signal using a graph-based transform, where the method includes: receiving a template index for a target unit from the video signal, wherein the template index indicates a graph-based transform template to be applied to the target unit; deriving a graph-based transform matrix for the target unit corresponding to the template index; and decoding the target unit based on the graph-based transform matrix. In some scenarios, this enables efficient compression of inter-predicted residual blocks by using graph-based transform templates. The graph-based transform template may enable adaptive transform selection, thus achieving improved coding gain. Further, by defining a graph-based transform template or a template index, coding with a low-complexity is enabled that can reduce the number of bits to be transmitted to the decoder.