Intra Prediction Reference Derivation for Unreconstructed Image Blocks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image encoding/decoding technologies face challenges with low prediction accuracy when neighboring reference samples are unavailable, particularly in high-resolution and high-quality images, leading to increased transmission and storage costs.
Innovation Solution
An improved image encoding/decoding method that derives reference samples for intra prediction using unreconstructed areas based on template matching, motion information, and neural networks, along with traditional reconstructed samples, to enhance prediction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If simple padding method is used to derive reference sample, then device complexity is reduced, but prediction accuracy deteriorates
Solution Approach 1:
The patent performs preliminary actions by deriving reference samples for unreconstructed areas using template matching and motion information before the actual intra prediction process. This allows the prediction algorithm to access more accurate reference samples earlier in the encoding/decoding pipeline, improving overall prediction accuracy without adding complex real-time processing during prediction.
Solution Approach 2:
The patent introduces an intermediary mechanism by using template matching and motion information as intermediate steps to derive reference samples. Instead of directly using simple padding, the system uses these intermediary techniques to generate more accurate reference samples from unreconstructed areas, which then serve as input for the intra prediction process.
2Measurement precision
If reference sample from unreconstructed area is derived using template matching, then prediction accuracy is improved, but processing time increases
Solution Approach 1:
Template matching and motion information derivation are performed as preliminary actions before the main intra prediction process. By preparing accurate reference samples in advance from unreconstructed areas, the system reduces the computational burden during the actual prediction phase, balancing accuracy improvement with processing efficiency.
Solution Approach 2:
The patent applies template matching and motion compensation selectively to derive reference samples only for unreconstructed areas where they are most needed, rather than applying these computationally intensive methods uniformly across all areas. This partial application approach maintains prediction accuracy where required while minimizing unnecessary processing time.
3Measurement precision
If high-resolution images are transmitted using existing media, then image quality is maintained, but transmission cost increases
Solution Approach 1:
The patent changes the parameter of reference sample quality by deriving more accurate reference samples from unreconstructed areas using template matching and motion information. This improvement in reference sample accuracy enables more efficient intra prediction, which reduces the number of bits required to encode high-resolution images, thereby lowering transmission costs while maintaining image quality.
Solution Approach 2:
The system uses feedback from motion information and template matching results to continuously improve reference sample derivation. By incorporating this feedback loop, the encoding/decoding process adapts to the actual content characteristics, achieving better compression efficiency for high-resolution images without sacrificing quality.
4Measurement precision
If neural network model is used to derive reference sample, then prediction accuracy is improved, but device complexity increases
Solution Approach 1:
The neural network model performs preliminary derivation of reference samples for unreconstructed areas before the main intra prediction process. By pre-processing and generating accurate reference samples using the neural network, the system reduces the complexity of the subsequent prediction stage, as the neural network handles the complex pattern recognition and sample generation tasks in advance.
Data Source
AI summary
An image encoding/decoding method and apparatus, a recording medium storing a bitstream and a transmission method are provided. The image decoding method may comprise determining an intra prediction mode of a current block, deriving a reference sample of an unreconstructed area, and performing intra prediction according to the intra prediction mode based on at least one of the derived reference sample of the unreconstructed area or a reference sample of a reconstructed area. The reference sample of the unreconstructed area may be a reference sample of an area located at at least one of a left bottom, bottom, right bottom, right or right top of the current block.