Video Encoding Orthogonal Transformer Hardware Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing moving-picture coding methods, such as H.264, require multiple hardware configurations for different prediction modes, leading to increased circuit scale and complexity due to the need for various orthogonal and inverse orthogonal transformation processes.
Innovation Solution
A moving-picture encoding and decoding apparatus that uses two types of orthogonal transformers, one optimized for intra-prediction direction and the other for discrete cosine transformation, reducing the number of required hardware configurations and circuit scale while improving coefficient density through predictive residual management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If multiple hardware configurations are used for different prediction modes, then encoding efficiency is improved, but circuit scale increases
Solution Approach 1:
The patent applies universality by designing a single orthogonal transformation unit that can handle multiple prediction modes (intra-prediction and inter-prediction) through software control rather than requiring separate hardware configurations for each mode. This multi-functional approach maintains encoding efficiency across different prediction types while using a unified hardware structure.
Solution Approach 2:
The patent uses copying by implementing transformation coefficients and prediction data in memory rather than duplicating hardware circuits. The same orthogonal transformation unit is reused across different prediction modes by loading different transformation matrices and parameters from memory, avoiding the need for multiple physical transformation units.
2Manufacturing precision
If orthogonal transformation processes are expanded for each prediction mode, then coefficient density is improved, but device complexity increases
Solution Approach 1:
The patent applies dynamics by making the orthogonal transformation process adaptable through software control. The transformation unit dynamically adjusts its operation based on the prediction mode being used, selecting appropriate transformation matrices and parameters from memory rather than requiring fixed hardware configurations for each mode. This allows coefficient density optimization for each mode while using a single flexible hardware unit.
Solution Approach 2:
The patent uses parameter changes by modifying transformation parameters (matrices, coefficients) based on prediction mode rather than changing hardware structure. Different prediction modes use different transformation parameters stored in memory, allowing the same hardware unit to achieve mode-specific coefficient density optimization through parameter adjustment rather than structural modification.
Data Source
AI summary
According to one embodiment, an encoding apparatus includes a prediction unit, a classifying unit, a first transformer, a second transformer, an order controller, and an entropy coder. The prediction unit obtains a predictive residual signal to be encoded, by using a mode selected from intra-prediction modes. The first transformer obtains first transformation coefficients by subjecting the signal to an orthogonal transformation by use of a first transformation basis if the selected mode is classified into a mode having a prediction direction. The first transformation basis is preset so that a coefficient density after the orthogonal transformation is higher than a coefficient density.