Image Ringing Artifact Reduction via Gradient-Based Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for reducing ringing artefacts in digital images, such as those used in JPEG and MPEG compression, face issues with misclassification due to threshold dependence on contrast levels and the complexity of order-statistical information, leading to unwanted artefacts like blurring or flickering.
Innovation Solution
A method that adjusts contrast by obtaining a blurred image using recursive spatially-adaptive edge-preserving filters and combining it with the original image based on ring likelihoods, which are calculated as the product of non-edge pixel likelihood and strongest edge likelihood, to reduce ringing artefacts while preserving edges.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If classification-based approaches with thresholds on local variance are used to reduce ringing artefacts, then ringing noise can be targeted, but misclassification occurs due to dependence on contrast level and lighting conditions
Solution Approach 1:
The patent changes the parameter basis from local variance (which depends on contrast and lighting) to gradient magnitude and orientation. This parameter transformation makes the classification independent of absolute brightness levels and contrast variations, thereby resolving the misclassification problem while maintaining the ability to identify edge pixels for ringing noise reduction
2Object-affected harmful factors
If order-statistical information is used in filtering operations, then ringing artefacts can be reduced, but the complexity increases and artefacts such as local flickering are created
Solution Approach 1:
The patent extracts only the essential gradient information (magnitude and orientation) from the image data, discarding the complex order-statistical information. This extraction approach maintains the ability to preserve edges and reduce ringing artefacts while significantly reducing computational complexity and avoiding the non-linear jumps that cause flickering artefacts
3Measurement precision
If training over a large database of images is performed to reduce misclassification rate, then classification accuracy improves, but the quality of obtained filters decreases and computational cost increases
Solution Approach 1:
The patent employs self-service by using the image's own gradient information to guide the filtering process. The gradient magnitude and orientation are computed directly from the image data itself, eliminating the need for external training databases. This self-adaptive approach maintains high classification accuracy while improving processing efficiency and filter quality
Data Source
Figure 1~2
Figure 3
Figure 4~6
AI summary
A method and apparatus is provided for adjusting the contrast of an input image (1). A blurred image (2) corresponding to the input image (1) is obtained, the brightness level of at least some of the pixels in the input image (1) being varied to provide the blurred image (2). A ring likelihood (4) for pixels in the input image (1) is obtained, the ring likelihood (4) providing a measure of the likelihood that said pixels are non-edge pixels in the neighbourhood of an edge pixel. The output (7) image is produced as a sum of the brightness levels in the input image (1) and the blurred image (2) in dependence on the ring likelihoods.