Image reconstruction with incomplete fourier-space magnitude data combined with real-space information
View Patent ↗Image reconstruction is based on phase retrieval by combining incomplete Fourier-space magnitude data with real-space information. Phase retrieval is performed based on the Fourier-space magnitude data, where the real-space information is expressed in a form suitable to use as a phase retrieval constraint, preferably using a wavelet-space representation. The use of incomplete Fourier-space magnitude data advantageously reduces the amount of data required compared to approaches that need comprehensive Fourier-space magnitude data. The real space information can be regarded as partial information of the image being reconstructed. Depending on the application, more or less real space information may be available.
1. A method of comparing an object to a corresponding object description, the method comprising:
a) providing a real-space representation R of said object based on said object description;
b) providing substantially incomplete Fourier-space magnitude data of said object by performing a diffraction measurement on said object;
c) computing a wavelet-space representation W of said real-space representation R;
d) dividing basis wavelets of said wavelet-space representation into a first set W 1 having coefficients a 1 and a second set W 2 having coefficients a 2 ;
e) computing a wavelet-space representation of an initial estimate;
f) iteratively refining said estimate one or more times to provide a final estimate, wherein coefficients of wavelets in said set W 1 are held fixed and coefficients of wavelets in said set W 2 are allowed to vary as agreement between said Fourier-space magnitude data and a Fourier-space representation of said estimate is sought; and
g) comparing said final estimate to said object description, and providing results of said comparison as an output.
2. The method of claim 1 , wherein every wavelet in said set W 1 has a scale greater than or equal to the scale of any wavelet in said set W 2 .
3. The method of claim 1 , wherein said iteratively refining said estimate comprises:
i) projecting said estimate onto a Fourier-space representation having Fourier magnitudes that agree with said Fourier-space magnitude data, at points where said Fourier-space magnitude data exists;
ii) projecting said estimate onto a wavelet-space representation having the same basis wavelets as said wavelet-space representation W and having coefficients equal to said coefficients a 1 for wavelets in said set W 1 ; iii) repeating said steps (i) and (ii) in succession one or more times to provide said final estimate.
4. The method of claim 1 , wherein said object and said object description are two-dimensional.
5. The method of claim 1 , wherein said object and said object description are three-dimensional.
6. The method of claim 1 , wherein said object is a fabricated integrated circuit, and wherein said object description is a design specification of said fabricated integrated circuit.
7. The method of claim 6 , wherein said Fourier-space magnitude data is obtained along Fourier-space axes that correspond to real-space axes aligned with traces of said fabricated integrated circuit.
8. The method of claim 1 , wherein said initial estimate is computed by setting said initial estimate equal to said representation W, except that coefficients of wavelets in said second set W 2 are set to zero.