Layered Segmentation Decoding for Efficient Video Bitstreams

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image and video coding technologies, including hybrid codecs and machine learning applications, face challenges in efficiently decoding data with scalability and adaptability to content characteristics, leading to suboptimal bitstream efficiency and processing time.

Innovation Solution

A method and apparatus for decoding data using a cascaded structure of segmentation information processing layers, which process sets of segmentation information elements in multiple layers, enabling efficient parsing and upsampling, and allowing for parallel processing on GPUs/NPUs, with trainable convolution kernels for improved decoding efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Ease of manufacture

If traditional hybrid coding methods are used with separate optimization of transformation, quantization, and entropy coding, then each component can be optimized independently, but the overall decoding efficiency and adaptability to content characteristics are suboptimal

Engineering Contradiction:
ImproveIndependent optimization of coding componentsVSAvoidDecoding efficiency
Core Design Contradiction:
Ease of manufactureVSProductivity

Solution Approach 1:

The patent merges traditionally separate coding components (transformation, quantization, entropy coding) into a unified neural network framework where segmentation information is processed through multiple cascaded layers that jointly optimize all components. This integration allows the system to achieve both independent component optimization and improved overall decoding efficiency simultaneously.

Inventive Principle:
Principle #5Merging (Combining)

Solution Approach 2:

The patent applies segmentation by dividing the decoding process into multiple processing layers that handle different aspects of segmentation information separately. Each layer processes specific segmentation elements independently before combining results, enabling both modular optimization and efficient parallel processing.

Inventive Principle:
Principle #1Segmentation

2Adaptability or versatility

If machine learning is applied to determine or optimize prediction parameters, then adaptability to content characteristics improves, but the amount of side information needed in the bitstream increases

Engineering Contradiction:
ImproveAdaptability to content characteristicsVSAvoidAmount of side information in bitstream
Core Design Contradiction:
Adaptability or versatilityVSQuantity of substance

Solution Approach 1:

The patent extracts only the essential segmentation information needed for adaptive decoding while discarding redundant data. By selectively transmitting only critical segmentation elements through the bitstream, the system maintains high adaptability to content characteristics while minimizing the quantity of side information required.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent changes the representation parameters of segmentation information by encoding it in a compressed hierarchical format. This parameter transformation reduces the bitstream requirements while preserving the adaptability needed for content-specific optimization, allowing the decoder to reconstruct detailed segmentation maps from compact representations.

Inventive Principle:
Principle #35Parameter changes

3Productivity

If segmentation information is processed through multiple cascaded layers with upsampling, then decoding adaptability and efficiency improve, but processing complexity increases

Engineering Contradiction:
ImproveDecoding efficiencyVSAvoidProcessing complexity of segmentation layers
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The patent resolves complexity issues by transitioning to parallel processing across multiple dimensions using GPU/NPU architectures. Instead of sequential processing through cascaded layers, the system processes segmentation information simultaneously across multiple processing units, maintaining high decoding efficiency while reducing the time dimension of complexity.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

Solution Approach 2:

The patent uses copying by replicating processing layers across multiple hardware instances in parallel. Each layer is copied to dedicated processing units that operate simultaneously, reducing overall processing time and distributing complexity across multiple identical modules rather than requiring a single complex sequential processor.

Inventive Principle:
Principle #26Copying

Data Source

PatentUS20260075231A1Decoding with signaling of segmentation information
Publication Date: 2026.03.12 HUAWEI TECH CO LTD
  • US20260075231A1 patent drawing
  • US20260075231A1 patent drawing
  • US20260075231A1 patent drawing

AI summary

The present disclosure relates to methods and apparatuses for decoding data for (still or video processing into a bitstream). Two or more sets of segmentation information elements are obtained from the bitstream. Then, each of the two or more sets of segmentation information elements are inputted respectively into two or more segmentation information processing layers out of a plurality of cascaded layers. In each of the two or more segmentation information processing layers, the respective sets of segmentation information are processed. The decoded data for picture or video processing are obtained based on the segmentation information processed by the plurality of cascaded layers. Accordingly, the data may be decoded from the bitstream in an efficient manner in the layered structure.