Hybrid input-output algorithm: Difference between revisions
m cat |
m WP:CHECKWIKI error fixes + general fixes, References after punctuation per WP:REFPUNC and WP:PAIC using AWB (7510) |
||
Line 5: | Line 5: | ||
Although it has been shown that the method of error reduction strictly converges to the correct or optimal solution <ref>Bauschke HH, Combettes PL, Luke DR,”Phase retrieval, error reduction algorithm, and Fienup variants: a view from convex optimization” jOURNAL OF THE OPTICAL SOCIETY OF AMERICA A-OPTICS IMAGE SCIENCE AND VISION,(2005) 19:1334-1345</ref> |
Although it has been shown that the method of error reduction strictly converges to the correct or optimal solution <ref>Bauschke HH, Combettes PL, Luke DR,”Phase retrieval, error reduction algorithm, and Fienup variants: a view from convex optimization” jOURNAL OF THE OPTICAL SOCIETY OF AMERICA A-OPTICS IMAGE SCIENCE AND VISION,(2005) 19:1334-1345</ref> |
||
<ref>Fienup J.R, “Reconstruction of an object from the modulus of its fourier transform” Optics letters (1978) 3:1,</ref> |
<ref>Fienup J.R, “Reconstruction of an object from the modulus of its fourier transform” Optics letters (1978) 3:1,</ref> |
||
there is no limit to how long this process can take. Moreover, the error reduction algorithm will almost certainly find a local minima instead of the global. The HIO differs from error reduction only in one step but this is enough to reduce this problem significantly. Whereas the error reduction approach iteratively improves solutions over time the HIO remodels the previous solution in Fourier space applying negative feedback. By minimizing the mean square error in the Fourier space from the previous solution, the HIO provides a better candidate solution for inverse transforming. Although being both faster and more powerful, the HIO does have a uniqueness problem |
there is no limit to how long this process can take. Moreover, the error reduction algorithm will almost certainly find a local minima instead of the global. The HIO differs from error reduction only in one step but this is enough to reduce this problem significantly. Whereas the error reduction approach iteratively improves solutions over time the HIO remodels the previous solution in Fourier space applying negative feedback. By minimizing the mean square error in the Fourier space from the previous solution, the HIO provides a better candidate solution for inverse transforming. Although being both faster and more powerful, the HIO does have a uniqueness problem.<ref>Miao J, Kirz J, Sayre D, “The oversampling phasing metod”, Acta Chryst. (2000), D56, 1312-1315</ref> |
||
Depending on how strong the negative feedback is there can more often be more than one solution for any set of diffraction data. This might seem like a big problem but it has been shown that many of these possible solutions stem from the fact that HIO allows for mirror images taken in any plane to arise as solutions. In crystallography, the scientist are seldom interested in the atomic coordinates relative to any other reference than the molecule itself and is therefore more than happy with a solution that is upside-down of flipped from the actual image. On the downside, HIO does have a tendency to be able to escape both global and local maxima. This is probably also depending on the strength of the feedback parameter and a good solution to this problem is to switch algorithm when the error reaches its minimum. Other methods of phasing a coherent diffraction pattern include [[difference map algorithm]] and "relaxed averaged alternating reflections" or RAAR.<ref>1.Luke Russel D, “Relaxed averaged alternating reflections for diffraction imaging” Inverse problems, (2005) 21, 37-50</ref> |
Depending on how strong the negative feedback is there can more often be more than one solution for any set of diffraction data. This might seem like a big problem but it has been shown that many of these possible solutions stem from the fact that HIO allows for mirror images taken in any plane to arise as solutions. In crystallography, the scientist are seldom interested in the atomic coordinates relative to any other reference than the molecule itself and is therefore more than happy with a solution that is upside-down of flipped from the actual image. On the downside, HIO does have a tendency to be able to escape both global and local maxima. This is probably also depending on the strength of the feedback parameter and a good solution to this problem is to switch algorithm when the error reaches its minimum. Other methods of phasing a coherent diffraction pattern include [[difference map algorithm]] and "relaxed averaged alternating reflections" or RAAR.<ref>1.Luke Russel D, “Relaxed averaged alternating reflections for diffraction imaging” Inverse problems, (2005) 21, 37-50</ref> |
||
==References== |
==References== |
||
<!--- See [[Wikipedia:Footnotes]] on how to create references using <ref></ref> tags which will then appear here automatically --> |
<!--- See [[Wikipedia:Footnotes]] on how to create references using <ref></ref> tags which will then appear here automatically --> |
||
{{Reflist}} |
{{Reflist}} |
||
{{DEFAULTSORT:Hybrid Input Output (Hio) Algorithm For Phase Retrieval}} |
{{DEFAULTSORT:Hybrid Input Output (Hio) Algorithm For Phase Retrieval}} |
Revision as of 19:05, 28 December 2010
This article needs additional citations for verification. (April 2010) |
Hybrid input-output (HIO) algorithm for phase retrievalis a modification of the error reduction algorithm for retriving the phases in Coherent diffraction imaging. Determining the phases of a diffraction pattern is crucial since the diffraction pattern of an object is its Fourier transform and in order to properly inverse transform the diffraction pattern the phases must be known. The amplitude however, can be measured from the intensity of the diffraction pattern and can thus be known experimentally. This fact together with some kind of support (mathematics) can be used in order to iteratively calculate the phases. The HIO algorithm uses negative feedback in fourier space in order to progressively force the solution to conform to the fourier domain constraints (support). Unlike the error reduction algorithm which alternately applies fourier and object constraints the HIO "skips" the object domain step and replaces it with negative feedback acting upon the previous solution.
Although it has been shown that the method of error reduction strictly converges to the correct or optimal solution [1] [2] there is no limit to how long this process can take. Moreover, the error reduction algorithm will almost certainly find a local minima instead of the global. The HIO differs from error reduction only in one step but this is enough to reduce this problem significantly. Whereas the error reduction approach iteratively improves solutions over time the HIO remodels the previous solution in Fourier space applying negative feedback. By minimizing the mean square error in the Fourier space from the previous solution, the HIO provides a better candidate solution for inverse transforming. Although being both faster and more powerful, the HIO does have a uniqueness problem.[3] Depending on how strong the negative feedback is there can more often be more than one solution for any set of diffraction data. This might seem like a big problem but it has been shown that many of these possible solutions stem from the fact that HIO allows for mirror images taken in any plane to arise as solutions. In crystallography, the scientist are seldom interested in the atomic coordinates relative to any other reference than the molecule itself and is therefore more than happy with a solution that is upside-down of flipped from the actual image. On the downside, HIO does have a tendency to be able to escape both global and local maxima. This is probably also depending on the strength of the feedback parameter and a good solution to this problem is to switch algorithm when the error reaches its minimum. Other methods of phasing a coherent diffraction pattern include difference map algorithm and "relaxed averaged alternating reflections" or RAAR.[4]
References
- ^ Bauschke HH, Combettes PL, Luke DR,”Phase retrieval, error reduction algorithm, and Fienup variants: a view from convex optimization” jOURNAL OF THE OPTICAL SOCIETY OF AMERICA A-OPTICS IMAGE SCIENCE AND VISION,(2005) 19:1334-1345
- ^ Fienup J.R, “Reconstruction of an object from the modulus of its fourier transform” Optics letters (1978) 3:1,
- ^ Miao J, Kirz J, Sayre D, “The oversampling phasing metod”, Acta Chryst. (2000), D56, 1312-1315
- ^ 1.Luke Russel D, “Relaxed averaged alternating reflections for diffraction imaging” Inverse problems, (2005) 21, 37-50