Image Encoding Quantization Matrix for Weighted Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing HEVC and VVC encoding methods do not effectively perform quantization control based on frequency components for weighted intra-/inter-prediction, leading to suboptimal subjective image quality.
Innovation Solution
An image encoding apparatus that uses a quantization matrix specifically designed for weighted intra-/inter-prediction, allowing for quantization processing that improves subjective image quality by controlling quantization for each frequency component.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a quantization matrix based on conventional prediction method is used, then the encoding process is simple, but quantization control according to frequency component cannot be performed for weighted intra-/inter-prediction
Solution Approach 1:
The patent applies local quality by creating a dedicated quantization matrix specifically for weighted intra-/inter-prediction modes, rather than using a universal quantization matrix for all prediction types. This allows frequency-component-specific quantization control to be applied locally to the prediction error signals of weighted prediction modes, improving quantization precision for this specific functionality without requiring complete redesign of the entire encoding system.
2Productivity
If weighted intra-/inter-prediction is implemented, then encoding efficiency is improved, but subjective image quality cannot be optimized due to lack of frequency-based quantization control
Solution Approach 1:
The patent applies parameter changes by modifying the quantization parameters (quantization matrix values) specifically for frequency components of prediction error signals in weighted intra-/inter-prediction. By adjusting quantization step values based on frequency characteristics, the system optimizes the balance between encoding efficiency and subjective image quality, allowing high-frequency components to be quantized more aggressively while preserving important low-frequency information.
3Adaptability or versatility
If quantization matrix is designed for conventional prediction, then compatibility with existing HEVC is maintained, but it does not correspond to new weighted intra-/inter-prediction method
Solution Approach 1:
The patent applies segmentation by separating the quantization matrix into different types: a first quantization matrix for conventional intra-prediction and a second quantization matrix for weighted intra-/inter-prediction. This segmentation allows the system to maintain compatibility with existing HEVC conventions for conventional prediction while adding support for the new weighted prediction method, enabling gradual adoption without requiring complete system replacement.
Data Source
AI summary
The present invention provides an image encoding apparatus for encoding a sequence of images, where the apparatus comprises a prediction unit which generates, for a target block to be encoded having a predetermined size in an image, a predicted image from both an intra-prediction image and an inter-prediction image, and obtains prediction errors that are differences between the target block and the predicted image; a transform unit which frequency-transforms the prediction errors, a quantization unit which quantize, using a quantization matrix, the transform coefficients, and an encoding unit which entropy-encodes quantized transform coefficients, wherein the quantization unit obtains the quantization matrix, using at least one of a predetermined quantization matrix for the intra-prediction and a predetermined quantization matrix for inter-prediction.