Image Encoding With Weighted Inter-Intra Synthesis for Code Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image encoding techniques using prediction images synthesized by averaging inter and intra prediction images are insufficient in reducing code amounts effectively.
Innovation Solution
An image encoding technique that generates a synthesis prediction by weighting and averaging inter and intra prediction images, allowing multiple types of intra predictions, and determining the weighting parameter based on the type of intra prediction and prediction modes of adjacent blocks.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If simple averaging synthesis of inter and intra prediction images is used, then device complexity is reduced, but code amount reduction is insufficient
Solution Approach 1:
The patent changes the parameter of synthesis processing from simple averaging to weighted averaging, where different weights are applied to inter prediction image and intra prediction image based on block characteristics. This allows better exploitation of complementary features between inter and intra predictions, improving code amount reduction while maintaining manageable device complexity through standardized weighting rules.
Solution Approach 2:
The patent introduces dynamic weighting where the synthesis method adapts to different block types and prediction scenarios. By selectively applying different weights to inter and intra prediction images based on block characteristics, the system dynamically optimizes the synthesis process for various content types, achieving better compression without excessive complexity.
2Manufacturing precision
If multiple types of intra predictions are used for synthesis processing, then image quality is improved, but device complexity increases
Solution Approach 1:
The patent applies local quality by using different intra prediction types (planar, angular, DC) for synthesis processing depending on the specific block characteristics and content type. Rather than uniformly applying one method, the system selects appropriate prediction types for different regions, improving overall image quality while keeping complexity manageable through localized optimization.
Solution Approach 2:
The patent segments the prediction processing into different intra prediction types (planar, angular, DC) that can be selectively applied to different blocks. This segmentation allows the system to choose the most appropriate prediction method for each block's characteristics, improving image quality without requiring all prediction types to be simultaneously complex.
Data Source
AI summary
An image encoding method for encoding an image, includes: a prediction image generation step generating a synthesis-prediction image by performing synthesis processing synthesizing an inter-prediction image and an intra-prediction image to an encoding-target block; and an encoding step encoding a difference between the prediction image generated in the prediction image generation step and a pixel value of an image of the encoding-target block, a plurality of intra prediction types are allowed to be used as the intra prediction used for the synthesis processing, and the plurality of intra prediction types include an in-screen block copy prediction, the synthesis processing includes weighting processing performed to the inter-prediction image and the intra-prediction image, and a weighting parameter of the intra-prediction image in the weighting processing is determined in accordance with an intra prediction type of the encoding-target block and prediction mode combination of a plurality of blocks adjacent to the encoding-target block.