Video Signal Approximation Transform Using Preprocessing Matrices
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current video processing technologies face challenges in efficiently handling next-generation video content with high spatial resolution, high frame rate, and high dimensionality, requiring more complex and resource-intensive transform designs for coding efficiency.
Innovation Solution
A method is introduced to approximate specific cosine or sine transforms by adding preprocessing and postprocessing steps to existing transforms, reducing complexity and memory requirements through the use of preprocessing and postprocessing matrices, and enabling selective application of these steps based on transform selection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a specific cosine transform or sine transform is used to achieve high coding efficiency, then coding efficiency is improved, but calculation complexity and memory requirements increase
Solution Approach 1:
The transform process is divided into three separate stages: preprocessing step (multiplication by preprocessing matrix), core transform step (cosine or sine transform), and postprocessing step (multiplication by postprocessing matrix). This segmentation allows each stage to be optimized independently, with the core transform using simplified algorithms while preprocessing and postprocessing handle the complexity of achieving high coding efficiency.
Solution Approach 2:
Preprocessing matrices and postprocessing matrices are introduced as intermediary components between the input signal and the core transform, and between the core transform and the output signal. These intermediary matrices enable the use of simpler core transforms while maintaining the coding efficiency of more complex transforms through the mathematical relationship: Target Transform ≈ Postprocessing Matrix × Core Transform × Preprocessing Matrix.
2Reliability
If a specific cosine transform or sine transform is used to achieve high coding efficiency, then coding efficiency is improved, but memory requirements increase
Solution Approach 1:
The preprocessing matrices and postprocessing matrices are designed to be shared across multiple transform types. By establishing mathematical relationships between different transforms (e.g., DST7, DCT8, DST6, flipped DST7), the same preprocessing and postprocessing matrices can serve multiple transform operations, significantly reducing the total memory required to store transform matrices.
Solution Approach 2:
The preprocessing and postprocessing matrices are designed with universal applicability, where a single set of matrices can be used for approximating multiple different target transforms. This multi-functionality reduces memory requirements by eliminating the need to store separate complete transform matrices for each transform type.
3Reliability
If next-generation video content with high spatial resolution and high frame rate is processed, then video quality is improved, but processing power requirements increase
Solution Approach 1:
By segmenting the transform into preprocessing, core transform, and postprocessing stages, the computationally intensive operations are distributed. The core transform uses simpler algorithms requiring less processing power, while preprocessing and postprocessing steps use efficient matrix multiplications, overall reducing the processing power burden for high-resolution, high-frame-rate video.
Solution Approach 2:
The invention changes the parameter representation of transforms by using approximation relationships. Instead of directly implementing complex transforms with high processing requirements, the system uses the parameter relationship: Target Transform Parameters ≈ Postprocessing Matrix × Core Transform Parameters × Preprocessing Matrix, enabling efficient computation while maintaining video quality.
Data Source
AI summary
The present invention provides a method for reconstructing a video signal by using an approximation transform on the basis of a preprocessing/postprocessing matrix, the method comprising: obtaining a transform index of a current block from a video signal, wherein the transform index corresponds to any one of a plurality of transform combinations including at least one of DST7, DCT8, DST6, or flipped DST7, and a transform combination includes a horizontal transform and a vertical transform; deriving a primary transform for horizontal/vertical directions corresponding to the transform index; performing an inverse primary approximation transform by using the primary transform for the horizontal/vertical directions, wherein the inverse primary approximation transform is generated by multiplying the primary transform for the horizontal/vertical directions by a preprocessing matrix and a postprocessing matrix; and reconstructing the video signal by using the inverse primary approximation-transformed current block.