Edge-Guided Demosaicing for Sharper Images and Fewer Color Artifacts
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing demosaicing methods in image sensors result in artifacts such as zipper effect, block noise, reduced image detail, false color, and blurring due to interpolation errors, particularly at edges, leading to a trade-off between noise filtering and sharpness.
Innovation Solution
A demosaicing method and device that includes a main edge estimator, a main edge enhancer, and a sub interpolator to improve edge detection and directionality, using high-frequency edge enhancement and gradient-based interpolation to reduce color errors and enhance image sharpness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If linear interpolation is used to restore missing color pixel values, then the demosaicing process is simple and fast, but image sharpness decreases and edges become blurred
Solution Approach 1:
The patent changes the parameter of interpolation method from simple linear interpolation to gradient-based directional interpolation. By calculating gradients and determining edge directions, the algorithm adapts the interpolation approach based on local image characteristics, preserving edge sharpness while maintaining computational feasibility
Solution Approach 2:
The patent introduces dynamic adaptation by estimating edge directions and adjusting interpolation strategies based on detected edge orientations. The algorithm dynamically selects interpolation directions perpendicular to edges, making the demosaicing process adaptive rather than static, thereby preserving image sharpness
2Object-affected harmful factors
If filtering is applied to reduce noise, then noise is reduced, but sharp edges become blurred
Solution Approach 1:
The patent applies different processing strategies to different local regions of the image. By detecting edge directions locally, the algorithm performs directional interpolation that preserves edge sharpness in edge regions while applying appropriate filtering in non-edge regions, achieving local optimization rather than uniform processing
Solution Approach 2:
The patent segments the image processing into distinct stages: gradient calculation, edge direction estimation, and directional interpolation. This segmentation allows the algorithm to selectively apply noise reduction and edge preservation techniques to different components, avoiding the trade-off between noise filtering and edge sharpness
3Productivity
If color channels are restored in sequence using previous step results, then the processing is efficient, but errors accumulate
Solution Approach 1:
The patent performs preliminary gradient calculations and edge direction estimations using the green channel data before interpolating red and blue channels. This preliminary action provides accurate directional information that guides the subsequent color channel restoration, preventing error accumulation by establishing a reliable foundation first
4Ease of operation
If averaging of pixels crossing edges is performed, then interpolation is simple, but image sharpness at edges decreases
Solution Approach 1:
Instead of averaging pixels across edges (which blurs edges), the patent inverts the approach by identifying edge directions first and then performing interpolation perpendicular to edges. This inversion prevents averaging across edge boundaries, thereby preserving edge sharpness while maintaining interpolation functionality
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A demosaicing method for an input mosaic image includes: obtaining an input image, for each sub target pixel, estimating a main edge direction based on main color pixel values; determining the main edge direction by performing edge enhancement based on the input mosaic image; generating a main color demosaiced image by performing a main interpolation based on the main edge direction to generate interpolated main color pixel values corresponding to the sub target pixels; and generating a first sub color demosaiced image and a second sub color demosaiced image by performing a sub interpolation to generate interpolated first sub color pixel values and interpolated second sub color pixel values for the sub target pixels and for main target pixels corresponding to the main color pixel values.