Non-Rigid HDR Image Registration via Unreliable Pixel Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing high-dynamic range (HDR) imaging techniques face challenges in registering and blending images captured at different exposure settings, leading to ghosting artifacts due to object motion, with current methods being either slow and robust or fast but inaccurate, failing to achieve interactive frame rates without sacrificing quality.
Innovation Solution
A system and method for fast, non-rigid registration of HDR image stacks, involving generating a warped image through sparse to dense flow field propagation, detecting unreliable pixels, and correcting them by blending with reference image pixels, using a patch match algorithm to ensure accurate geometric consistency and exposure adjustment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional robust registration techniques are used, then registration accuracy is improved, but processing speed deteriorates (cannot achieve interactive frame rates)
Solution Approach 1:
The patent segments the registration process into multiple stages: feature detection, sparse matching, dense flow field propagation, and pixel reliability classification. This segmentation allows each stage to be optimized independently, achieving both accuracy and speed by processing only critical pixels in detail while handling remaining pixels more efficiently.
Solution Approach 2:
The patent applies local quality by identifying and separately processing unreliable pixels versus reliable pixels. Unreliable pixels (those prone to ghosting artifacts) receive intensive correction processing, while reliable pixels are processed more simply. This localized approach concentrates computational resources where they are most needed, improving overall accuracy without proportionally increasing total processing time.
2Productivity
If fast registration techniques are used, then processing speed is improved (interactive frame rates achieved), but registration accuracy deteriorates (visible image artifacts remain)
Solution Approach 1:
The patent introduces an intermediary classification step that identifies unreliable pixels before final blending. This intermediary process acts as a mediator between fast processing and accurate results by directing full correction processing only to pixels that need it, while allowing reliable pixels to be processed quickly without intensive correction.
Solution Approach 2:
The patent applies partial action by performing complete correction processing only on a subset of pixels identified as unreliable, rather than processing all pixels with the same intensive method. This partial approach achieves sufficient accuracy for the critical pixels while maintaining overall processing speed, avoiding the excessive computation that would result from uniformly processing all pixels at maximum detail.
3Productivity
If images are blended without motion adjustment, then processing speed is improved, but image quality deteriorates (ghosting artifacts occur)
Solution Approach 1:
The patent performs preliminary action by detecting and correcting unreliable pixels before final image blending. This advance identification and correction of problematic pixels prevents ghosting artifacts from occurring during the blending process, eliminating the need for post-processing artifact removal while maintaining processing efficiency.
Solution Approach 2:
The patent applies preliminary anti-action by specifically targeting and correcting pixels that would otherwise create ghosting artifacts. By identifying unreliable pixels and applying correction processing before blending, the system preemptively counteracts the formation of ghosting artifacts, preventing the harmful effect rather than remedying it afterward.
Data Source
AI summary
A system, method, and computer program product are provided for performing fast, non-rigid registration for at least two images of a high-dynamic range image stack. The method includes the steps of generating a warped image based on a set of corresponding pixels, analyzing the warped image to detect unreliable pixels in the warped image, and generating a corrected pixel value for each unreliable pixel in the warped image. The set of corresponding pixels includes a plurality of pixels in a source image, each pixel in the plurality of pixels associated with a potential feature in the source image and paired with a corresponding pixel in a reference image that substantially matches the pixel in the source image.