Motion-Blurred HDR Image Deblurring With Adaptive MAP Priors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing deblurring techniques struggle to accurately recover items of interest in motion-blurred images, particularly in high-dynamic range (HDR) images, especially when bright point sources are present, leading to noise contamination and poor target detection.
Innovation Solution
A method using successive approximations and adaptive Maximum A Posteriori (MAP) deblurring with L1 and Total Variation (TV) priors, where bright pixels are deblurred first, followed by iterative refinement using MAP estimates with adaptive priors, and optionally converting to polar coordinates for rotational motion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional deblurring techniques are applied to motion-blurred HDR images, then processing speed is maintained, but deblurring accuracy deteriorates due to noise contamination and inability to handle bright point sources
Solution Approach 1:
The patent applies different prior models to different regions of the image: L1 prior for bright point source regions and TV prior for smooth target regions. This local differentiation allows the deblurring process to handle various image characteristics optimally, improving deblurring accuracy while preventing noise contamination in different regions
Solution Approach 2:
The patent performs successive approximations by first identifying and deblurring bright point sources, then removing them from the image before processing smooth targets. This preliminary action prevents noise from bright sources from contaminating the subsequent target detection process
2Measurement precision
If adaptive priors are used to improve deblurring accuracy, then target recovery improves, but computational complexity increases
Solution Approach 1:
The patent segments the deblurring process into distinct stages: first processing bright point sources with L1 prior, then processing smooth targets with TV prior. This segmentation allows each prior to be applied only where needed, improving target recovery accuracy while managing computational complexity through divided processing
Solution Approach 2:
The patent removes bright point sources from the image after deblurring them, so they do not participate in subsequent processing. This partial action reduces the computational burden on the TV prior processing stage while maintaining high accuracy for both point sources and targets
3Adaptability or versatility
If rotational motion deblurring is applied to handle nonlinear motion, then motion complexity is addressed, but processing difficulty increases due to complex blurring kernels
Solution Approach 1:
The patent transforms the image to polar coordinates locally, where rotational motion appears as radial smearing that can be handled more easily. This local coordinate transformation simplifies the blurring kernel structure for rotational motion while maintaining adaptability to handle both linear and nonlinear motion types
Data Source
AI summary
A method of deblurring an image taken by a moving imaging sensor. The method includes: receiving an image from the imaging sensor; collecting motion data indicative of motion of the sensor while capturing the image; applying successive approximations in deblurring based on certain features of the image; and applying a Maximum A Posteriori (MAP) deblurring process with adaptive priors to deblur the image and create a deblurred image, wherein the MAP deblurring process applies a first prior in certain regions and second prior in other regions.