Method and system for wavefront reconstruction

• US 4,933,872 A
• Filed: 11/15/1988
• Issued: 06/12/1990
• Est. Priority Date: 11/15/1988
• Status: Expired due to Fees

• Current Assignee: Eastman Kodak Company
• Original Assignee: Eastman Kodak Company
• Inventors: Antognini, Thomas C., Weaver, John C., Liff, Harold J., Vandenberg, Donald E.
• Primary Examiner(s): Smith, Jerry
• Assistant Examiner(s): Gordon, Paul
• Application Number:
  US07/271,802
• Time in Patent Office: 574 Days
• Field of Search: 364/513, 356/121, 356/354, 250/201, 250/553, 250/578, 382/14, 382/15, 382/29
• US Class Current: 706/22
• CPC Class Codes: G01J 9/00 Measuring optical phase difference devices or arrangements for controlling the phase of light beams G02F1/01; Determining degree of coherence; Measuring optical wavelength spectrometry G01J3/00
• CPC Class Codes: G02B 27/0025 for optical correction, e.g. distorsion, aberration
• CPC Class Codes: G06N 3/04 Architecture, e.g. interconnection topology
• CPC Class Codes: G06T 2207/20084 Artificial neural networks [ANN]
• CPC Class Codes: G06T 5/60 using machine learning, e.g. neural networks
• CPC Class Codes: G06T 5/73 Deblurring; Sharpening
• CPC Class Codes: G06V 10/42 Global feature extraction by analysis of the whole pattern, e.g. using frequency domain transformations or autocorrelation
• CPC Class Codes: G16H 50/20 for computer-aided diagnosis, e.g. based on medical expert systems
• CPC Class Codes: Y10S 706/912 Manufacturing or machine, e.g. agricultural machinery, machine tool
• CPC Class Codes: Y10S 706/924 Medical

Abstract

Method and system for wavefront reconstruction from an image plane intensity distribution profile. An imaging device may be an agent for producing the image plane intensity distribution profile, for example, a point spread function. In one embodiment, the method and system include defining a feature vector from the point spread function, and employing an adaptive computational architecture for associating the feature vector with at least one identifying characteristic of the imaging device, e.g., such as an amount of astigmatism.

Claims

1. A method to be employed in conjunction with a radiation system, the radiation system comprising:

(a) at least one source of radiation;

(b) at least one imaging device for imaging at least a portion of the radiation emitted by the or each source onto at least one image plane;

the or each imaging device functioning to produce at the image plane an intensity distribution profile; and

(c) a detector array for recording the intensity distribution profile at the image plane;

the method comprising the steps of;

(1) defining a feature vector derived from the intensity distribution profile; and

(2) employing an adaptive computational architecture for mapping the feature vector to at least one identifying characteristic of a selected imaging device.

2. A method according to claim 1, comprising deriving the feature vector from a point spread function produced by a point source of radiation.

3. A method according to claim 1, wherein the intensity distribution profile is recorded from an extended source for deriving the feature vector comprising power spectral components.

4. A method according to claim 1, comprising deriving the feature vector from a multi-dimensional moment vector.

5. A method according to claim 4, wherein the moment vector is defined by the equation

6. A method according to claim 1, comprising learning steps of deriving the feature vector by:

(a) inserting into the radiation system a set of n different imaging devices, each imaging device having known aberrations, for yielding n unique point spread functions; and

(b) developing a multi-dimensional moment vector for each of the n unique point spread functions.

7. A method according to claim 1, comprising deriving the feature vector from an energy spectra.

8. A method according to claim 1, wherein the step of employing the adaptive computational architecture comprises:

(a) providing a preliminary learning mode; and

(b) providing a subsequent real-time processing mode.

9. A method according to claim 8, comprising providing a non-linear neural network computational architecture.

10. A method according to claim 9, wherein the learning mode comprises the steps of:

iteratively adjusting a set of weighting factors defined by the neural network, by associating a succesion of known feature vectors with a succession of known aberrations.

11. A method according to claim 10, wherein the real time processing mode comprises the step of:

characterizing a heretofore arbitrary feature vector, by processing the arbitrary feature vector through the neural network, the characterizing, at least in part, based on the learning mode.

12. A method according to claim 8, wherein the adaptive computational architecture comprises providing a statistical learning algorithm.

13. A method according to claim 12, comprising the learning mode steps of:

(a) locating a known feature vector in a multi-dimensional feature vector space, the vector space corresponding to terms of orthogonal basis vectors; and

(b) developing a region in the vector space, the region defined by the known feature vector and statistical noise components.

14. A method according to claim 13, comprising the real-time processing mode step of:

characterizing a heretofore arbitrary feature vector by determining its location with respect to the region.

15. A radiation system comprising:

(a) at least one source of radiation;

(b) at least one imaging device for imaging at least a portion of the radiation emitted by the or each source onto at least one image plane;

the or each imaging device functioning to produce at the image plane an intensity distribution profile;

(c) a detector array for recording the intensity distribution profile at the image plane;

(d) means for defining a feature vector derived from the intensity distribution profile; and

(e) means for mapping the feature vector to at least one identifying characteristic of a selected imaging device.

16. A radiation system according to claim 15, wherein the source of radiation comprises a single point source emitting monochromatic radiation.

17. A radiation system according to claim 15, wherein the imaging device comprises a mirror.

18. A radiation system according to claim 15, wherein the imaging device comprises a lens.

19. A radiation system according to claim 15, wherein the detector array realizes a point spread function.

20. A radiation system according to claim 15, wherein the means for defining a feature vector comprises a means for deriving a moment analysis of a point spread function.

21. A radiation system according to claim 15, wherein the means for associating the feature vector with at least one identifying characteristic of a selected imaging device comprises a neural network, the neural network comprising:

(a) a plurality of column input nodes, for inputting a column feature vector;

(b) a plurality of non-linear column hidden nodes, each of which nodes inputs adjustable, weighted input signals w_ij from each of the input nodes;

(c) a plurality of column output nodes, each of which nodes accepts adjustable weighted input signals w_kl from each of the hidden nodes; and

(d) a plurality of column target nodes that are mated to the column output nodes.

22. A radiation system according to claim 21, wherein each of the non-linear column hidden nodes comprises:

means for operating on the nodes inputs in accordance with a pair of equations (1) and (2);

space="preserve" listing-type="equation">B.sub.i =f(Σ

w.sub.ij z.sub.j) (1)

space="preserve" listing-type="equation">f(x)=1/(1+exp (-x)) (2)

where,

w_ij =a set of weighted signals;

z_j =contents of an input node number j;

x=Σ

w_ij z_j ; and

f(x)=a non-linear discriminator.

23. A radiation system according to claim 21, wherein the neural network comprises:

(a) means for operating in a preliminary learning mode by iteratively adjusting the set of weighting signals w_ij and w_kl, in response to associating a succession of known feature vectors with a succession of known imaging devices; and

(b) means for operating in a subsequent real-time processing mode by processing an arbitrary feature vector through the neural network, and characterizing the arbitrary feature vector, at least in part, based on the learning mode.