Shape-adaptive codec decomposes alpha planes for image compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current shape coding methods for video sequences, particularly in scenarios like TV broadcasting and movies, fail to effectively utilize structural redundancies in matte and soft segmentation images, leading to inefficient compression and visual artifacts, especially in high-resolution images.
Innovation Solution
A shape-adaptive model-based codec that decomposes binary alpha planes into deterministic and stochastic components, allowing for flexible lossless and lossy compression modes, and uses rotated micro-processing units for improved compression ratios, while supporting semi-transparent objects and overlapping foreground and background elements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If block-based methods with fixed-size blocks are used for shape coding, then the coding process is simple and regular, but the compression ratio is limited due to redundant blocks that do not contain contour information
Solution Approach 1:
The shape coding process is segmented into two distinct phases: contour coding phase (using chain code for boundary pixels) and filling phase (using block-based methods for interior pixels). This segmentation allows each phase to use the most appropriate coding method, achieving both high compression ratio and coding simplicity where needed.
Solution Approach 2:
Different coding methods are applied to different regions of the shape: chain code is used for the contour boundary where precision is critical, while block-based filling is used for the interior regions where redundancy is high. This local adaptation optimizes the overall compression ratio while maintaining shape accuracy.
2Loss of information
If contour-based methods are used for shape coding, then the compression ratio is improved, but substantial pre-processing is required and the coding complexity increases
Solution Approach 1:
The contour is traced and identified in advance during a preliminary contour coding phase, allowing the main filling phase to proceed efficiently using simple block-based methods. This preliminary action separates the complex contour extraction task from the simpler filling task, reducing overall coding complexity.
Solution Approach 2:
The contour tracing process continuously identifies boundary pixels and generates chain code representation, which then directly feeds into the filling phase. This continuous workflow eliminates the need for separate pre-processing steps and maintains coding efficiency throughout the process.
3Device complexity
If fixed-size blocks are used in block-based methods, then the coding structure is regular and simple, but many blocks become redundant and consume storage space without contributing to shape representation
Solution Approach 1:
Instead of applying block-based filling to all blocks within the bounding box, the method performs partial action by only filling blocks that are identified as interior blocks (those completely inside the contour). This eliminates redundant blocks that would otherwise consume storage space without contributing to shape representation.
Solution Approach 2:
The method extracts and identifies the contour boundary first, then uses this information to extract only the necessary interior blocks for filling. Blocks outside the contour or on the boundary are excluded from the filling process, removing redundancy and improving storage efficiency.
4Measurement precision
If lossless compression is used for binary shapes, then the shape accuracy is preserved, but the bit rate requirement is high and compression efficiency is reduced
Solution Approach 1:
The method changes the coding parameters differently for boundary pixels versus interior pixels: boundary pixels use chain code with higher precision to maintain shape accuracy, while interior pixels use block-based filling with lower bit rate requirements. This parameter differentiation achieves both shape accuracy and bit rate efficiency.
Solution Approach 2:
The shape is segmented into boundary pixels and interior pixels, with each segment processed using the most efficient compression method. This segmentation allows lossless compression for the critical boundary while using more efficient lossy block-based filling for the interior, improving overall bit rate efficiency.
Data Source
Figure 1(a)~2
Figure 3~4
Figure 5~6
AI summary
The present invention relate to methods and codecs for image and video compression. Embodiments of the present invention include a novel shape-adaptive model-based codec (SAM) that supports binary shapes as well as matte and soft segmentation image compression by decomposing input shapes into deterministic and stochastic components for flexible lossy and lossless coding. The present invention can provide inter/intra prediction and flexibly adapts between lossy and lossless modes with various parameters for compression quality control. The compression module can also be adapted with numerous other compression techniques.