Chromatic Aberration Correction via Spectral Mapping Model
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image processing technologies lack an effective chromatic aberration correction method, leading to deviations in color representation between acquired images and true object colors due to differences in spectral tristimulus values between imaging systems and human observation, as well as gamma correction deviations.
Innovation Solution
A model training method is developed to construct a mapping model between original and standard color information values using a coefficient matrix, with an objective function based on a regular term of a variable exponent, allowing for the determination of the coefficient matrix through iteration and convergence, and applying this to correct chromatic aberration in target images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If chromatic aberration correction is performed using conventional methods, then color correction is attempted, but the correction accuracy is insufficient due to spectral tristimulus value differences between imaging systems and human observation
Solution Approach 1:
The patent transforms the chromatic aberration correction problem from conventional color space transformations to a spectral domain problem. By converting color information to spectral tristimulus values and using variable exponent regularization to optimize the mapping between imaging system responses and human visual system responses, the method achieves more accurate color correction that accounts for fundamental spectral differences between systems.
Solution Approach 2:
The patent introduces spectral tristimulus values as an intermediary representation between the imaging system's raw color data and the final corrected color output. This intermediary spectral domain allows for more accurate modeling of the transformation between different color observation systems, bridging the gap between machine capture and human perception.
2Measurement precision
If gamma correction deviation is corrected, then display accuracy is improved, but the overall chromatic aberration correction remains insufficient due to multiple affecting factors
Solution Approach 1:
The patent merges gamma correction with chromatic aberration correction into a unified spectral domain processing framework. Rather than treating these as separate correction steps, the method simultaneously optimizes both transformations through the spectral tristimulus value mapping, reducing overall system complexity while improving comprehensive color accuracy.
3Reliability
If a mapping model with coefficient matrix is constructed, then the relationship between original and standard color values is established, but the model stability is affected by the choice of regularization term
Solution Approach 1:
The patent employs variable exponent regularization instead of fixed exponent regularization in the objective function. This dynamic approach allows the regularization strength to adapt locally across different spectral components, optimizing the trade-off between model stability and mapping precision for each region of the spectral domain rather than applying a uniform constraint.
Data Source
AI summary
A model training method includes: constructing a mapping model between original color information values and standard color information values of images with a coefficient matrix; establishing an objective function of the mapping model based on a regular term of a variable exponent; determining original color information values and standard color information values of a plurality of sample images; and calculating the objective function from the original color information values and the standard color information values of the plurality of sample images to determine the coefficient matrix.