Linear Partitioning for Image Intra Coding Prediction Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The precision of predictive coding in digital image processing, particularly in high spatial activity images, is not optimized due to the distance between pixels to be predicted and their reference pixels being too large, leading to inefficiencies in compression and transmission.
Innovation Solution
The method involves partitioning macroblocks into non-square or rectangular shapes and using specific traversal orders for prediction, allowing for more localized frequency transforms and utilizing reference pixels from adjacent, previously coded partitions to refine prediction accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If macroblocks are partitioned into smaller rectangular or square blocks for spatial prediction, then the coding structure is simplified and processing is easier, but the prediction accuracy deteriorates because reference pixels are located far from the pixels to be predicted
Solution Approach 1:
The macroblock is segmented into multiple linear partitions (rows or columns) instead of traditional rectangular blocks. This segmentation allows prediction to proceed sequentially from one linear partition to the next, ensuring that reference pixels are always adjacent to the pixels being predicted, thereby resolving the contradiction between structured processing and prediction accuracy.
Solution Approach 2:
The patent changes the partitioning dimension from traditional 2D rectangular blocks to 1D linear partitions (either horizontal rows or vertical columns). This dimensional change enables the prediction process to maintain spatial locality while preserving a systematic processing structure, thus improving prediction accuracy without sacrificing coding organization.
2Device complexity
If reference pixels are taken from distant macroblocks for prediction, then the coding process is simpler with fewer constraints, but the prediction accuracy deteriorates especially in high spatial activity images
Solution Approach 1:
The patent applies local quality by making the prediction process dependent on the local spatial context. By using adjacent linear partitions as reference, the prediction adapts to local image characteristics (such as edges and details in high spatial activity regions), thereby improving prediction accuracy without significantly increasing overall process complexity.
3Ease of manufacture
If traditional square or rectangular macroblock partitioning is used, then the coding method is simpler and more standardized, but the adaptation to particular image structures or patterns is reduced
Solution Approach 1:
The patent introduces dynamics by allowing the partitioning orientation to be selected adaptively based on image content characteristics. The system can dynamically choose between horizontal and vertical linear partitions, and even switch between different traversal orders, thereby adapting to various image structures while maintaining a relatively simple coding framework.
Solution Approach 2:
The patent changes the partitioning parameter from fixed rectangular shapes to variable linear partitions with different orientations and traversal orders. This parameter change enables better adaptation to specific image patterns (such as horizontal or vertical edges) while keeping the overall coding method straightforward and implementable.
Data Source
Figure 1~2
Figure 3A~4F
Figure 5A~8
AI summary
The invention relates to a method of coding an image or a sequence of images generating a stream (F) of data comprising data representative of at least one group of pixels in one of said images (IE). Such a method comprises the steps of: - selecting (C1), for a group of pixels to be coded, an initial partition of predetermined linear form, - slicing (C2) said group of pixels to be coded according to said selected initial partition, - propagating (C2) said initial partition selected in said group of pixels to be coded, until the latter is sliced in its entirety into a plurality of partitions of predetermined form, - selecting, for said sliced group of pixels, a predetermined order of traversal of said partitions, - predicting and coding (C4) said partitions one after the other, according to said selected order of traversal.