Image Noise Reduction via Alignment Weighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image noise reduction methods face challenges in aligning multiple images captured in quick succession, particularly when there is camera movement, leading to geometric misalignment and motion blur, which affects the sharpness of the combined image.
Innovation Solution
A method using multiple kernel tracking and Lucas Kanade Inverse algorithms to determine the positions of candidate regions and apply transformations to align images, discarding flat regions and using error metrics to refine the alignment, thereby reducing noise and misalignment artifacts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple images are combined to reduce noise, then noise reduction is achieved, but geometric misalignment and motion blur are introduced
Solution Approach 1:
The patent applies preliminary alignment actions by computing transformation parameters (homography matrices) for each image in the sequence before combining them. This preliminary geometric correction ensures that images are properly aligned in terms of perspective and position before the noise reduction combining process, preventing misalignment artifacts in the final image.
Solution Approach 2:
The patent employs feedback mechanisms by using feature matching (SIFT, SURF, or ORB) to detect correspondences between images, computing transformation parameters based on these correspondences, and then applying transformations to align images. This closed-loop approach continuously refines alignment quality through iterative feature detection, transformation computation, and validation.
2Measurement precision
If images are captured in quick succession for noise reduction, then noise reduction effectiveness increases, but motion blur and camera movement increase
Solution Approach 1:
The patent applies preliminary motion compensation by computing transformation parameters that account for camera movement and scene motion before combining images. This preliminary correction aligns moving objects across frames and compensates for camera shake, preventing motion blur artifacts while maintaining the temporal proximity needed for effective noise reduction.
Solution Approach 2:
The patent applies local quality principles by using feature-based transformation that can handle different motion characteristics in different regions of the image. By matching local features (SIFT, SURF, ORB) and computing local transformations, the system can preserve sharpness in regions with significant motion while maintaining noise reduction benefits in stable regions.
3Manufacturing precision
If transformation parameters are computed to align images, then geometric alignment is improved, but processing complexity increases
Solution Approach 1:
The patent applies segmentation by dividing the alignment process into distinct stages: feature detection, feature matching, transformation parameter computation, and image transformation. This segmented approach allows each stage to be optimized independently and enables parallel processing of multiple images, reducing overall processing complexity while maintaining high alignment precision.
Solution Approach 2:
The patent employs parameter changes by using different transformation models (affine, perspective, or Euclidean) based on the specific alignment requirements and computational resources available. The system can adjust the complexity of transformation parameters computed (homography matrices with different degrees of freedom) to balance alignment quality with processing complexity.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
A reduced noise image can be formed from a set of images. One of the images of the set can be selected to be a reference image and other images of the set are transformed such that they are better aligned with the reference image. A measure of the alignment of each image with the reference image is determined. At least some of the transformed images can then be combined using weights which depend on the alignment of the transformed image with the reference image to thereby form the reduced noise image. By weighting the images according to their alignment with the reference image the effects of misalignment between the images in the combined image are reduced. Furthermore, motion correction may be applied to the reduced noise image.