Non-Rectangular Video Block Partitioning for Curved Boundary Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video compression techniques using straight-line segment partitioning fail to efficiently follow object boundaries, resulting in higher prediction errors and reduced compression efficiency, especially for blocks with curved boundaries.
Innovation Solution
Implementing exponential partitioning that divides rectangular blocks into non-rectangular regions using curved lines, characterized by exponential functions or predefined templates, allowing for improved alignment with object boundaries and reducing residual errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If straight-line segment partitioning is used, then the partitioning method is simple and easy to implement, but it fails to efficiently follow object boundaries resulting in higher prediction errors
Solution Approach 1:
The patent applies curvature by replacing straight-line partition boundaries with exponential curve boundaries. The exponential partitioning uses functions of the form y = a * exp(bx) to create curved boundaries that better conform to object shapes, particularly for regions with curved boundaries. This resolves the contradiction by introducing curvature to improve prediction accuracy while maintaining reasonable implementation complexity through predefined templates and parameters.
Solution Approach 2:
The patent changes the mathematical parameters of the partition boundary from linear equations to exponential equations. By using exponential functions with adjustable parameters (a, b, c, d) and predefined templates, the system can adapt the boundary shape to match object contours more accurately. This parameter change enables better following of curved object boundaries while keeping the implementation manageable through template-based approaches.
2Measurement precision
If exponential partitioning with curved lines is used, then alignment with object boundaries is improved and prediction errors are reduced, but the complexity of the partitioning method increases
Solution Approach 1:
The patent segments the exponential partitioning approach into predefined templates and parameter sets. Instead of requiring complex real-time curve fitting, the system divides the boundary representation into discrete templates (e.g., 8 templates with different curvature characteristics) and selects appropriate parameters from these segments. This segmentation reduces implementation complexity while maintaining the ability to represent curved boundaries accurately.
Solution Approach 2:
The patent uses copying by creating predefined templates of exponential curve segments that can be reused across different partitioning scenarios. These templates represent common boundary shapes and can be copied and applied to different regions, reducing the need for complex calculations in each case. The templates serve as reusable patterns that simplify the implementation while maintaining prediction accuracy.
3Productivity
If non-rectangular partitioning is used, then compression efficiency is enhanced by lowering residuals, but the complexity of encoding and decoding algorithms increases
Solution Approach 1:
The patent applies curvature to the partition boundaries using exponential functions, which better match the natural curved boundaries of objects in video content. This improved boundary alignment reduces the residual error between predicted and actual pixel values, enhancing compression efficiency. The curved boundaries allow for more accurate region-of-interest identification and motion compensation.
Solution Approach 2:
The patent changes from rectangular to exponential partitioning by introducing exponential function parameters (a, b, c, d) and template indices. These parameter changes enable flexible boundary representation that adapts to different object shapes. The use of predefined templates with specific parameter sets balances the need for accurate boundary representation with the constraint of algorithmic complexity.
Data Source
AI summary
An encoder for generating an encoded bitstream for decoding by a decoder includes a circuitry configured to receive a bitstream including a coded picture having a coding unit with at least one non-straight, non-rectangular boundary therein, in which decoding includes: determining start and end points of the boundary using first and second index information provided in the bitstream, generating predictive pixel values in the regions on each side of the boundary, smoothing the predictive pixel values across the boundary, and adding residual pixel values to the smoothed predictive pixel values.