Image Inpainting for Compression Artifact Correction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Existing image and video processing technologies suffer from artifacts caused by video compression, particularly around objects, and misalignment of depth information with object boundaries, leading to decreased image quality and interference with subsequent processing tasks.

Innovation Solution

A method that performs foreground/background segmentation based on depth or disparity information, creates an area around segmentation transitions, and applies foreground/background-dependent processing to correct artifacts by inpainting background or foreground information, thereby reducing compression artifacts and depth misalignment.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If lossy video compression schemes are used, then video compression efficiency is improved, but compression artifacts are introduced around objects

Engineering Contradiction:
Improvevideo compression efficiencyVSAvoidcompression artifacts
Core Design Contradiction:
ProductivityVSObject-affected harmful factors

Solution Approach 1:

The patent extracts and isolates the problematic regions (areas around objects) where compression artifacts occur. By segmenting the image into foreground and background based on depth information, the system identifies specific regions that require special processing to remove artifacts while preserving compression efficiency for the overall video stream.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent applies different processing qualities to different regions of the image. Instead of uniformly processing the entire image, it applies enhanced artifact removal only to local regions around objects where artifacts are present, while maintaining standard compression for other areas, thus balancing quality and efficiency.

Inventive Principle:
Principle #3Local quality

2Measurement precision

If semi-automated depth creation through annotation-and-propagation is used, then depth map generation is improved, but depth misalignment with object boundaries occurs

Engineering Contradiction:
Improvedepth map generationVSAvoiddepth alignment with object boundaries
Core Design Contradiction:
Measurement precisionVSManufacturing precision

Solution Approach 1:

The patent uses feedback from the segmented image and depth information to iteratively refine the depth map. By analyzing the segmentation transitions and comparing them with the depth data, the system identifies misalignments and corrects them through inpainting techniques, progressively improving depth accuracy.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The patent performs preliminary segmentation based on depth information before performing the actual depth correction. This preliminary action identifies which regions require depth adjustment, allowing the subsequent inpainting process to target specific areas and improve alignment efficiency.

Inventive Principle:
Principle #10Preliminary action

3Object-affected harmful factors

If inpainting techniques are applied to correct artifacts, then image quality is improved, but processing complexity increases

Engineering Contradiction:
Improveimage qualityVSAvoidprocessing complexity
Core Design Contradiction:
Object-affected harmful factorsVSDevice complexity

Solution Approach 1:

The patent segments the image into foreground and background regions based on depth information before applying inpainting. This segmentation allows the complex inpainting process to be divided into manageable regions, each processed independently with appropriate techniques, reducing overall computational complexity while maintaining quality.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent applies inpainting only to the extent necessary to correct artifacts in specific regions rather than processing the entire image uniformly. By targeting only the problematic areas around objects, the system reduces processing complexity and computational load while achieving the desired quality improvement.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP2316109B1Use of inpainting techniques for image correction
Publication Date: 2017.12.13 KONINKLIJKE PHILIPS NV
  • EP2316109B1 patent drawingFigure 1a~1b
  • EP2316109B1 patent drawingFigure 2~3
  • EP2316109B1 patent drawingFigure 4a~4b

AI summary

A method of processing an image signal comprising image and depth information is provided. The method is configured to perform segmentation on an image based on depth/disparity information present in the image signal comprising said image, and subsequently inpaint background for correction of the errors in the image around the foreground objects into a region that extends beyond the segment boundary of the foreground object and/or inpaint foreground for correction of errors in the image into a region that extends inside the segment boundary of the foreground object. In this way compression and other artifacts may be reduced.