Under-Display Camera Image Correction via Bayesian PSF Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image correction technologies, particularly for under-display cameras, face challenges in accurately deblurring images due to spatially varying point spread functions (PSFs) and noise, which limits the dynamic range and introduces artifacts like flare and residual glare, especially in high-contrast scenes.
Innovation Solution
The method employs a Bayesian approach to estimate the corruption operator in real-time, accounting for statistical distributions of images and noise, allowing for the correction of non-stationary PSFs and nonlinear color space corruption, and updates the corruption operator based on real-time image data to address gradual changes in the camera system.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional deblurring methods are used, then processing speed is maintained, but image quality deteriorates due to spatially varying PSFs and noise
Solution Approach 1:
The patent segments the image into multiple local regions and applies region-specific PSFs for deconvolution. This local segmentation approach allows the method to handle spatially varying PSFs effectively, improving image quality by adapting to local optical characteristics rather than using a single global PSF.
Solution Approach 2:
The patent implements dynamic PSF estimation that adapts to changing imaging conditions in real-time. The corruption operator is continuously updated based on current image data, allowing the system to handle temporal variations in optical properties and maintain high image quality under varying conditions.
2Reliability
If static PSF measurement is used, then manufacturing complexity is reduced, but reliability deteriorates due to gradual changes in camera system
Solution Approach 1:
The patent implements a feedback mechanism where the corruption operator is continuously updated based on real-time image data. This adaptive update process allows the system to compensate for gradual changes in camera components (such as lens degradation or sensor drift) by adjusting the PSF estimation to match current optical conditions, thereby maintaining high correction accuracy.
Solution Approach 2:
The patent performs preliminary PSF measurement during manufacturing to establish an initial corruption operator. This preliminary action provides a baseline that can be refined through real-time updates, reducing the immediate computational burden while ensuring reliability through subsequent adaptive adjustments.
3Measurement precision
If simple correction methods are used, then processing time is reduced, but image quality deteriorates in high-contrast scenes
Solution Approach 1:
The patent applies local quality by using region-specific PSFs tailored to different areas of the image. Each local region undergoes deconvolution with its own optimized PSF, which accounts for spatial variations in optical properties. This localized approach significantly improves image quality in high-contrast scenes compared to global methods, while the computational efficiency is maintained through efficient local processing.
4Measurement precision
If noise is not accounted for, then processing complexity is reduced, but measurement precision deteriorates
Solution Approach 1:
The patent incorporates noise characteristics by modeling them as statistical parameters within the Bayesian framework. The corruption operator estimation includes noise parameters that are jointly optimized with the PSF, allowing the system to distinguish between actual image features and noise artifacts. This parameter-based noise modeling improves measurement precision without requiring complex filtering operations.
Data Source
AI summary
In one embodiment, a method includes accessing (1) a corrupted image of a scene captured by a camera, (2) an estimated true image of the scene, (3) an estimated corruption operator f for the camera, and (4) one or more uncertainty metrics for f. The method further includes generating, by applying a corruption operation to the estimated true image and the corruption operator f, a predicted corrupted image of the scene captured by the camera and determining a difference between the predicted corrupted image and the corrupted image captured by the camera. The method further includes determining, based on the one or more uncertainty metrics for f, a likelihood distribution for the corruption operator f, and updating, based on the likelihood distribution for the corruption operator f and on the determined difference between the predicted corrupted image and the corrupted image captured by the camera, the estimated corruption operator f.