Banding Artefact Predictor Using Gradient Profile Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Lossy compression techniques used in image rendering often introduce visual artefacts such as banding artefacts, which are difficult to detect and eliminate, especially in compressed images and videos, affecting their quality and rendering efficiency.
Innovation Solution
A method is developed to detect banding artefacts by generating a gradient profile, candidate banding pixel map, banding edge map, and banding visibility map, using a Sobel operator and threshold values, and applying debanding filters to improve image quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If lossy compression techniques are applied to reduce image data size, then rendering speed and storage efficiency are improved, but visual artefacts such as banding artefacts are introduced
Solution Approach 1:
The patent applies debanding filters during the transcoding process before final image rendering. By detecting banding artefacts in the compressed video frames and applying smoothing filters to affected regions, the system prevents banding artefacts from appearing in the final rendered image, thus resolving the contradiction between compression efficiency and visual quality
Solution Approach 2:
The patent dynamically adjusts compression parameters and filter strength based on detected banding artefact severity. By analyzing gradient profiles and banding metrics of compressed images, the system modifies rendering parameters to reduce banding artefacts in critical regions while maintaining overall compression efficiency
2Quantity of substance
If higher compression levels are applied to reduce data size, then storage and transmission efficiency are improved, but the number and visibility of visual artefacts increase
Solution Approach 1:
The patent implements a feedback mechanism where banding artefacts are detected in compressed video frames, and the detection results are used to adjust subsequent processing steps. The system calculates banding metrics from gradient profiles and uses this feedback to apply targeted debanding filters, thereby reducing visual artefacts while maintaining compression efficiency
Solution Approach 2:
The patent applies debanding filters selectively to regions where banding artefacts are detected rather than uniformly across the entire image. By identifying affected regions through gradient analysis and applying localized smoothing, the system reduces visual artefacts in critical areas while preserving image quality and maintaining overall data efficiency
3Object-affected harmful factors
If banding artefacts are detected and debanding filters are applied, then visual quality is improved, but processing complexity and computational resources increase
Solution Approach 1:
The patent applies debanding filters only to regions where banding artefacts are detected above a certain threshold, rather than processing the entire image uniformly. By partially applying the filtering operation to affected regions identified through gradient profile analysis, the system improves visual quality while reducing overall processing complexity compared to full-image filtering
Solution Approach 2:
The patent segments the image processing task into distinct stages: gradient profile generation, banding detection, region identification, and selective filtering. By dividing the complex debanding process into manageable segments that operate on specific image regions, the system reduces processing complexity while maintaining effectiveness in removing banding artefacts
Data Source
AI summary
A method of measuring a banding artefact in an image includes generating a gradient profile from the image, where the gradient profile includes respective gradient magnitudes of pixels of the image; generating, using the gradient profile, a candidate banding pixel (CBP) map, where each location of the CBP map is such that a gradient magnitude of the gradient profile of a corresponding pixel of the image being greater than a first threshold and smaller than a second threshold; generating, using the CBP map, a banding edge map (BEM), where the BEM includes connected banding edges of the image; generating, using the BEM, a banding visibility map (BVM), where the BVM includes a respective banding metric for at least some pixels of the image; and generating a banding index of the image using the BVM.