Phase Unwrapping Segmentation for Interferometry
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional phase unwrapping methods in interferometry are time-consuming and noise-sensitive, struggling to accurately remove ambiguity in wrapped phase images due to the inherent ambiguity of the arctangent function and the presence of noise in input data.
Innovation Solution
The method involves segmenting the wrapped phase image into multiple regions based on phase values and locations, assigning integer values to represent phase offsets of 2π, and constructing an output phase image using neural networks for classification and post-processing to improve accuracy and robustness against noise.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional phase unwrapping methods are used, then phase ambiguity can be removed, but the process is time-consuming and noise-sensitive
Solution Approach 1:
The patent divides the phase image into multiple segments or regions based on phase discontinuities and characteristics. By segmenting the image, the algorithm can process each segment independently with optimized parameters, reducing overall computational time while maintaining accuracy. This segmentation approach allows faster processing compared to conventional methods that treat the entire image uniformly.
Solution Approach 2:
The patent applies different unwrapping strategies and parameters to different regions of the phase image based on their local characteristics. Regions with high phase discontinuities or noise are handled differently from smooth regions, allowing the algorithm to optimize processing for each local area and reduce overall processing time without sacrificing accuracy in critical regions.
2Measurement precision
If conventional phase unwrapping methods are used, then phase ambiguity can be removed, but the methods are noise-sensitive and produce errors
Solution Approach 1:
The patent performs preliminary processing steps before phase unwrapping, such as noise filtering, phase correction, and discontinuity detection. By preparing the phase image in advance and addressing noise issues before the main unwrapping process, the algorithm reduces noise sensitivity and produces more accurate results without requiring complex noise-resistant unwrapping algorithms.
Solution Approach 2:
The patent incorporates feedback mechanisms that continuously monitor the unwrapping process and adjust parameters or reprocess regions where errors are detected. This feedback loop allows the algorithm to correct noise-induced errors dynamically, improving overall accuracy and reducing sensitivity to noise in the input phase image.
3Loss of information
If the arctangent function is used for phase extraction, then phase information can be obtained, but ambiguity is introduced due to the periodic nature of the function
Solution Approach 1:
The patent transforms the one-dimensional phase values (0 to 2π) into a two-dimensional representation by introducing a winding number or branch index dimension. This additional dimension allows the algorithm to distinguish between different phase branches that the arctangent function cannot differentiate, resolving the ambiguity while maintaining complete phase information.
Solution Approach 2:
The patent segments the phase image into different branches or regions based on phase discontinuities and uses separate processing for each segment. By dividing the continuous phase field into discrete segments, the algorithm can track phase continuity across segments and resolve ambiguities that would otherwise be introduced by the periodic arctangent function.
Data Source
AI summary
Methods, apparatus and systems for processing interferograms in metrology applications are described. In one example aspect, a method includes obtaining an input phase image based on the interferograms, segmenting the input phase image by classifying the input phase image into multiple regions based on the phase value and a location of each pixel, assigning an integer value to each of the multiple regions, and constructing an output phase image based on the input phase image and the phase offset of each of the multiple regions.