Image Decoding with Low-Resolution Residual Group Reconstruction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional deep learning-based image compression schemes face high computational complexity due to serial or wave-front approaches in mean prediction, particularly with increasing feature resolution.

Innovation Solution

The method involves extracting image residual data, performing spatial resolution reduction to obtain extended residual groups, and conducting residual restoration on these groups at a lower resolution, followed by spatial resolution amplification to reconstruct the image.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If serial or wave-front approaches are used for mean prediction in deep learning-based image compression, then spatial redundancy reduction is achieved, but computational complexity increases significantly

Engineering Contradiction:
Improvespatial redundancy reductionVSAvoidcomputational complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent divides the feature map into multiple extended residual groups by partitioning the spatial dimensions. Each group is processed independently through residual restoration, which reduces the computational burden compared to processing the entire feature map sequentially. This segmentation allows parallel processing and reduces the overall computational complexity while maintaining spatial redundancy reduction effectiveness.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent introduces an extended residual group structure that reorganizes features along additional dimensions. By transforming the traditional serial processing approach into a multi-dimensional group processing approach, the method reduces computational complexity through dimensionality transformation while preserving the spatial redundancy reduction capability.

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

2Manufacturing precision

If feature resolution is increased to improve image quality, then reconstruction accuracy improves, but time complexity increases

Engineering Contradiction:
Improvereconstruction accuracyVSAvoidtime complexity
Core Design Contradiction:
Manufacturing precisionVSLoss of time

Solution Approach 1:

The patent segments the high-resolution feature map into multiple extended residual groups with reduced spatial resolution. Each group is processed at this lower resolution, which significantly reduces the time complexity. After processing, the groups are reconstructed back to the original resolution, maintaining high reconstruction accuracy while reducing computational time.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent changes the resolution parameter of the feature groups during processing. By operating at a reduced resolution for the residual restoration step and then restoring to the original resolution, the method achieves both high reconstruction accuracy and reduced time complexity. This parameter transformation is key to resolving the contradiction between accuracy and time.

Inventive Principle:
Principle #35Parameter changes

3Manufacturing precision

If residual restoration is performed on entire feature groups at original resolution, then reconstruction quality is maintained, but computational efficiency decreases

Engineering Contradiction:
Improvereconstruction qualityVSAvoidcomputational efficiency
Core Design Contradiction:
Manufacturing precisionVSProductivity

Solution Approach 1:

The patent segments the feature map into extended residual groups and performs residual restoration on each group at a reduced spatial resolution. This segmentation enables parallel processing and significantly improves computational efficiency. The groups are then reconstructed to the original resolution, maintaining reconstruction quality while improving productivity.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent reorganizes features into extended residual groups with different spatial dimensions. By processing in this reorganized dimension at lower resolution and then transforming back, the method achieves both high reconstruction quality and improved computational efficiency simultaneously.

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

Data Source

PatentEP4694118A1Image decoding method and apparatus, image coding method and apparatus, and device and storage medium
Publication Date: 2026.02.11 HANGZHOU HIKVISION DIGITAL TECHNOLOGY CO LTD
  • EP4694118A1 patent drawingFigure 1~2
  • EP4694118A1 patent drawingFigure 3
  • EP4694118A1 patent drawingFigure 4

AI summary

The present disclosure belongs to the field of image processing technologies, and in particular, to an image decoding method and apparatus, an image coding method and apparatus, a device and a storage medium. The image decoding method of the present disclosure includes: extracting image residual data or extended residual data from an image bitstream, and obtaining a plurality of extended residual groups based on the extracted image residual data or extended residual data; obtaining respective image reconstruction features corresponding to the extended residual groups by performing residual restoration on each of the plurality of extended residual groups; obtaining reconstructed feature data by performing spatial resolution amplification processing on the respective image reconstruction features corresponding to the extended residual groups; and obtaining a reconstructed image block by performing image reconstruction according to the reconstructed feature data. Since the obtained extended residual data has undergone spatial resolution reduction processing, residual restoration processing can be performed on entire groups at a lower resolution, thereby improving the overall computational efficiency of residual restoration and reducing time complexity.