Low-Frequency Shading Removal in Image Sensors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional scene-based non-uniformity correction methods for images are inadequate, particularly for fixed mount cameras and situations with partial motion, as they introduce artifacts and are not suitable for all camera configurations, leading to incomplete removal of low-frequency shading.
Innovation Solution
A method and system that generate a non-uniformity estimate using iterative addition of 2D Gaussian functions or other bump-like functions to approximate and remove low-frequency shading from images, applying filters to smoothen the estimates and update them based on a cost function that minimizes total variation and gradient energy, allowing for continuous image processing without interruption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If scene-based non-uniformity correction methods are used, then low-frequency shading can be corrected, but artifacts such as burn-in are introduced and the method is not suitable for fixed mount cameras
Solution Approach 1:
The patent segments the non-uniformity correction process into two independent components: a scene-based component that uses global motion information when available, and a temporal component that uses frame-to-frame differences. This segmentation allows the system to apply the appropriate correction method based on camera motion status, avoiding artifacts in fixed mount scenarios while maintaining correction accuracy when motion is present.
Solution Approach 2:
The patent implements a dynamic correction system that adapts its behavior based on detected global motion. When global motion is detected, the scene-based correction is activated; when no global motion is detected (fixed mount scenario), the temporal correction method is used instead. This dynamic adaptation eliminates the need for manual configuration and prevents artifact introduction in fixed mount scenarios.
2Adaptability or versatility
If scene-based correction methods relying on global motion are used, then non-uniformity can be corrected, but the method fails for fixed mount cameras and partial frame motion situations
Solution Approach 1:
The patent creates a universal non-uniformity correction system that can handle multiple camera configurations (fixed mount, moving camera, partial frame motion) through a single integrated approach. The system universally applies temporal analysis to detect global motion and selectively combines scene-based and temporal correction methods, making it adaptable to all camera types without requiring separate correction algorithms for each scenario.
3Measurement precision
If conventional non-uniformity correction methods are applied, then shading removal can be achieved, but the processing interrupts the image or video recording process
Solution Approach 1:
The patent performs preliminary analysis of frame-to-frame differences and global motion detection on previously captured frames before processing the current frame for non-uniformity correction. This preliminary action allows the system to determine the appropriate correction strategy in advance, enabling continuous processing without interruption to the recording pipeline.
Solution Approach 2:
The patent implements continuous non-uniformity correction that operates in parallel with the image recording process. By using temporal information from consecutive frames and performing corrections in a continuous manner rather than batch processing, the system maintains uninterrupted image or video recording while still achieving effective shading removal.
Data Source
Figure 1
Figure 2
Figure 3A~3C
AI summary
Various techniques are disclosed for separating and removing low-frequency shadow or shading (also referred to herein as "non-uniformity") from images that have been corrupted by the non-uniformity. A non-uniformity estimate that approximates the non-uniformity effect on the corrupted image may be generated by iteratively adding new blotches of non-uniformity data represented by two-dimensional (2D) functions, such as 2D Gaussian functions, to the non-uniformity estimate and applying filters to smoothen the 2D functions. In each iteration of the non-uniformity estimate generation process, a new non-uniformity update candidate that minimizes a cost function is identified. The corrupted image is processed based on the non-uniformity estimate to generate a corrected image.