Image Processing With Adaptive Blending Against Compression Artefacts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing techniques face a trade-off between temporal denoising and the introduction of compression artefacts when using lossy compression algorithms for reducing memory bandwidth and storage space, which degrades image quality.
Innovation Solution
A method and apparatus that modify blending coefficients based on identified image features to balance the contributions of new and accumulated image data, reducing noise while minimizing compression artefacts by adjusting the relative weights of pixel intensity values in the output image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If lossy compression algorithm is applied to reduce memory bandwidth and storage space, then memory bandwidth and storage space are reduced, but compression artefacts are introduced which degrade image quality
Solution Approach 1:
The patent modifies blending coefficients dynamically based on image features (such as motion magnitude, gradient magnitude, and pixel variance) to adjust the weight given to accumulated frame data versus current frame data. This parameter adjustment allows the system to maintain image quality in regions sensitive to compression artifacts while still utilizing compressed accumulated frame data in less sensitive regions, thus resolving the contradiction between compression efficiency and image quality.
2Reliability
If accumulated image data from multiple frames is used for temporal denoising, then noise is reduced, but compression artefacts are introduced due to lossy compression of the accumulated data
Solution Approach 1:
The patent applies different blending strategies to different regions of the image based on local image features. In regions with high motion or high gradient magnitude (where compression artifacts are more noticeable), the blending coefficient is adjusted to reduce reliance on compressed accumulated data. In regions with low motion and low gradient (where noise reduction is more beneficial and artifacts are less noticeable), the original blending coefficients are maintained. This local differentiation resolves the contradiction by optimizing both noise reduction and artifact minimization in their respective regions.
Solution Approach 2:
The system calculates image features (motion magnitude, gradient magnitude, pixel variance) from the compressed accumulated frame data and uses these features as feedback to dynamically adjust blending coefficients. This feedback mechanism allows the system to identify regions where compression artifacts are likely to occur and adjust the blending accordingly, maintaining noise reduction benefits while minimizing artifact propagation.
3Manufacturing precision
If blending coefficients are modified based on image features, then compression artefacts are minimized and image quality is improved, but additional processing steps are required
Solution Approach 1:
The patent calculates image features (motion magnitude, gradient magnitude, pixel variance) from the compressed accumulated frame data before performing the blending operation. By preparing these feature calculations in advance, the system avoids complex real-time analysis during the blending stage and enables efficient coefficient adjustment. This preliminary action reduces the overall processing complexity while maintaining image quality improvements.
Data Source
AI summary
A method and apparatus for processing image data is provided. The method comprises receiving decompressed accumulated image data that has been subjected to a lossy compression algorithm. The accumulated image data includes an accumulated frame of image data and a corresponding plurality of blending coefficients. The blending coefficients are updated by identifying an image feature associated with at least one pixel location of the accumulated frame of image data. The updated blending coefficients and decompressed accumulated frame of image data are sent to a temporal noise reducer, which is configured to use the updated blending coefficients to combine the decompressed accumulated frame of image data with a newly received frame of image data to generate an output image.