Multi-Scale Banding Detection for Accurate Image Artifact Identification

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Conventional banding artifact detection techniques in images and videos suffer from high false positives and false negatives, failing to accurately identify visible banding artifacts and incorrectly identifying edges or segments as artifacts.

Innovation Solution

Generate a plurality of banding confidence maps based on intensity difference values and image scales, using pre-processing operations like bit-depth conversion and low-pass filtering, to create a banding index that quantifies visible banding.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If false edge detection or false segment detection techniques are used to detect banding artifacts, then the detection process can be implemented, but the detection accuracy is low with high false positives and false negatives

Engineering Contradiction:
Improvebanding artifact detection accuracyVSAvoiddetection reliability
Core Design Contradiction:
Measurement precisionVSReliability

Solution Approach 1:

The patent segments the detection process into multiple independent components: gradient calculation, banding confidence map generation for each scale, and aggregation. By dividing the image into multiple scales and generating separate confidence maps for each scale, the method achieves more reliable detection with reduced false positives and negatives compared to single-scale approaches.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent extends the detection from a single scale to multiple scales by generating confidence maps at different resolutions. This multi-scale approach adds a dimensional aspect to the detection process, allowing the system to capture banding artifacts that may be visible at certain scales but not others, thereby improving overall detection accuracy and reliability.

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

2Measurement precision

If edge detection operations are performed to identify edges in source and processed images, then potential banding artifacts can be identified, but edges corresponding to actual image features may be incorrectly identified as banding artifacts

Engineering Contradiction:
Improvebanding artifact identification accuracyVSAvoidfalse positive detections
Core Design Contradiction:
Measurement precisionVSObject-generated harmful factors

Solution Approach 1:

The patent incorporates feedback mechanisms by comparing gradient magnitudes and directions across multiple scales. The banding confidence is calculated based on the consistency of gradient patterns across scales, providing a feedback loop that helps distinguish actual banding artifacts from legitimate image edges, thereby reducing false positive detections.

Inventive Principle:
Principle #23Feedback

Solution Approach 2:

The patent changes the detection parameters by operating at multiple scales rather than a single fixed scale. By adjusting the scale parameter and generating confidence maps at different resolutions, the system can adaptively identify banding artifacts while ignoring edges that represent actual image features, thus reducing false positives.

Inventive Principle:
Principle #35Parameter changes

3Measurement precision

If segmentation operations are performed on source and processed images to identify segments, then potential banding artifacts can be identified, but segments corresponding to actual image regions may be incorrectly identified as banding artifacts

Engineering Contradiction:
Improvebanding artifact identification accuracyVSAvoidfalse positive detections
Core Design Contradiction:
Measurement precisionVSObject-generated harmful factors

Solution Approach 1:

The patent segments the detection process into multiple independent components: gradient calculation, banding confidence map generation for each scale, and aggregation. By dividing the image into multiple scales and generating separate confidence maps for each scale, the method achieves more reliable detection with reduced false positives and negatives compared to single-scale approaches.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The patent changes the detection parameters by operating at multiple scales rather than a single fixed scale. By adjusting the scale parameter and generating confidence maps at different resolutions, the system can adaptively identify banding artifacts while ignoring edges that represent actual image features, thus reducing false positives.

Inventive Principle:
Principle #35Parameter changes

4Measurement precision

If detection operations are performed on pixels with very similar values, then banding artifacts may be detected, but edge detection operations may not successfully detect edges around the banding artifact

Engineering Contradiction:
Improvebanding artifact detection accuracyVSAvoidedge detection difficulty
Core Design Contradiction:
Measurement precisionVSDifficulty of detecting and measuring

Solution Approach 1:

The patent extends the detection from a single scale to multiple scales by generating confidence maps at different resolutions. This multi-scale approach adds a dimensional aspect to the detection process, allowing the system to capture banding artifacts that may be visible at certain scales but not others, thereby improving overall detection accuracy and reliability.

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

Data Source

PatentEP4285317B1Banding artifact detection in images and videos
Publication Date: 2026.03.04 NETFLIX INC
  • EP4285317B1 patent drawingFigure 1
  • EP4285317B1 patent drawingFigure 2
  • EP4285317B1 patent drawingFigure 3

AI summary

In various embodiments, a banding detection application generates a first set of pixel confidence values based on a first intensity difference value and first image scale associated with a first image, wherein each pixel confidence value included in the first set of pixel confidence values indicates a likelihood that a corresponding pixel included in the first image at the first image scale corresponds to banding in the first image. The banding detection application then generates a banding index corresponding to the first image based on the first set of pixel confidence values.