Method and system for reconstructing sampled signals

Patent metadata

Field	Value
Publication number	US-9510787-B2
Application number	US-201414567015-A
Country	US
Kind code	B2
Filing date	Dec 11, 2014
Priority date	Dec 11, 2014
Publication date	Dec 6, 2016
Grant date	Dec 6, 2016

Abstract

A method reconstructs a signal by sampling the signal using a sampling procedure to obtain an input signal. A consistent set is determined from the input signal including the first elements such that applying the sampling procedure to the first elements results in the input signal. According to the type of the signal, a guiding set is determined including second elements disjoint from the first elements. A reconstruction set including third elements is generated so that the third elements minimize a sum of a first similarity measure of the third elements to the second elements and a second similarity measure of the third elements to the first elements. A transformed signal that minimizes a function on the reconstruction set is determined. A reconstructed signal is rendered so that a third similarity measure of the reconstructed signal to the transformed signal is smaller than a tolerance.

First claim

We claim:

1. A method for reconstructing a signal, wherein the signal is a light field, comprising steps of:
sampling, using a sampling procedure, the signal to obtain an input signal, wherein the input signal is associated with a type;
determining a consistent set from the input signal, wherein the consistent set includes first elements such that applying the sampling procedure to the first elements results in the input signal;
determining, according to the type, a guiding set, wherein the guiding set includes second elements disjoint from the first elements;
generating a reconstruction set, wherein the reconstruction set includes third elements, wherein the third elements minimize a sum of a first similarity measure of the third elements to the second elements and a second similarity measures of the third elements to the first elements;
determining a transformed signal that minimizes a function on the reconstruction set;
generating a reconstructed signal so that a third similarity measure of the reconstructed signal to the transformed signal is smaller than a tolerance; and
rendering the reconstructed signal, wherein the input signal, the consistent set, the guiding set, the reconstruction set, the transformed signal and the reconstruction signal are stored in a memory connected to a processor performing the steps.

2. The method of claim 1 , wherein the type of the input signal is selected from a group consisting of an image, a patch in an image, a set of images, a video sequence, a patch in a video sequence, a depth map, a spectral map, a patch of a depth map, a patch of a map, image-related feature vectors, and combinations thereof, and further comprising:
determining the guiding set, wherein the determining is selected from a group consisting a Fourier transform, a cosine transform, a wavelet transform, a method for learning from a training dataset, and a spectral transform of a matrix of a graph and combinations thereof, and further comprising:
constructing the graph according to the type; and
determining the matrix of the graph using the graph and the input signal.

3. The method of claim 1 , wherein the input signal is streaming, further comprising steps of:
generating the reconstructed signal in real time.

4. The method of claim 1 , wherein one or a combination of the type, the consistent set, and the guiding set is determined using a first set of probability distributions, and further comprising:
determining the reconstruction set using a second set of probability distribution using a statistical similarity measure between stochastic random variables according to the first set of the probability distributions.

5. The method of claim 1 , wherein one or a combination of the consistent set and the guiding set comprises one or a group of first linear sets, and further comprising:
determining the reconstruction set using a second linear set according one or the group of the first linear sets.

6. The method of claim 5 , wherein the guiding set is a Krylov-based subspace selected from a group consisting of an approximate Krylov subspace, a rational Krylov subspace, an approximate rational Krylov subspace, and combinations thereof, and further comprising:
determining the reconstruction signal using a Krylov-based method according to the Krylov-based subspace.

7. The method of claim 5 , further comprising:
determining the guiding set using a range of a guiding transformation, comprising one or a combination of a projector, a frame function, and a filter function.

8. The method of claim 7 , further comprising:
approximating the filter function using one or a combination of a matrix function or an approximate matrix function, wherein the matrix function is determined using a group consisting of a matrix polynomial, an approximate matrix polynomial, a matrix rational function, an approximate matrix rational function, and combinations thereof.

9. The method of claim 8 , further comprising:
determining the matrix polynomial using a Chebyshev iterative method using one or a combination of roots of Chebyshev polynomials and recursive formulas for the Chebyshev polynomials.

10. The method of claim 1 , further comprising:
determining at least one of the first, second, and third similarity measures using one or a combination of a distance, a norm, a semi-norm, a correlation, a coherence, a divergence, and a metric.

11. The method of claim 1 , further comprising:
determining the function using the first similarity measure, the second similarity measure, and a parameter, such that a ratio of the first similarity measure to the second similarity measure equals the parameter at a minimum of the function.

12. The method of claim 11 , wherein the sampling procedure causes an inaccuracy in the input signal due to one or a combination of noise, a limited accuracy of a sensor used by the sampling procedure, and a limited precision of the input signal, further comprising:
determining a level of the inaccuracy in the input signal relative to a diameter of the reconstruction set; and
determining the parameter to be proportional to the level of the inaccuracy.

13. The method of claim 1 , wherein the reconstruction set is an interval with a first endpoint in the consistent set, and a second endpoint in the guiding set, and further comprising:
determining the first endpoint using one or a combination of solving a first system of equations and projecting the second endpoint to the consistent set;
determining the second endpoint using one or a combination of solving a second system of equations and projecting the first endpoint to the guiding set;
determining the reconstruction set as a convex linear combination of the first endpoint and the second endpoint; and
solving one or a combination of the first system of equations and the second system of linear equations using an iterative method.

14. The method of claim 13 , wherein the transformed signal is determined using an iterative method that minimizes the function, and further comprising:
adjusting a termination criteria of the iterative method using a training dataset; and
determining the iterative method using a conjugate gradient iterative method, a Chebyshev iterative method, a preconditioned iterative method, and combinations thereof.

15. The method of claim 2 , wherein the signal further includes a sound field and wherein the type of the input signal includes one or a set of audio sequences, patches in audio sequences, audio spectral maps, patches of audio spectral maps, audio feature vectors, and combinations thereof.

16. An apparatus for reconstructing a signal, wherein the signal is a light field, comprising:
at least one sensor for sampling the signal with a sampling procedure to obtain an input signal, wherein the input signal is associated with a type;
a memory for storing the input signal, a consistent set determined from the input signal, wherein the consistent set includes first elements such that applying the sampling procedure to the first elements of the consistent set results in the input signal, a guiding set of second elements disjoint from the first elements, determined according to the type, a reconstruction set including third elements, wherein the third elements minimize a sum of a first similarity measure of the third elements to the second elements set and a second similarity measures of the third elements to the first elements, a transformed signal that minimizes a function on the reconstruction set, a reconstructed signal so that a third similarity measure of the reconstructed signal to the transformed

Assignees

• Mitsubishi Electric Res Laboratories Inc

Inventors

• Kniazev Andrei
• Gadde Akshay
• Tian Dong
• Mansour Hassan
• Waters Richard C

Classifications

• G06F17/17 Primary

  Function evaluation by approximation methods, e.g. inter- or extrapolation, smoothing, least mean square method ({G06F17/18 takes precedence } ; interpolation for numerical control G05B19/18) · CPC title

• A61B5/72 Primary

  Signal processing specially adapted for physiological signals or for diagnostic purposes · CPC title

• G06F2218/22

  Source localisation; Inverse modelling · CPC title

• G06T12/00

  Tomographic reconstruction from projections · CPC title

• H04N19/00

  Methods or arrangements for coding, decoding, compressing or decompressing digital video signals · CPC title