Image restoration from transformed component data
AssigneeRegents of the University of Minnesota
InventorHal H. Ottesen
Life of this patent
Estimating the term…
Abstract
An image in a base domain (e.g., spatial) is restored from only one component (e.g., magnitude) of a transform-domain (e.g., Fourier) image by combining this component with a complementary component (e.g., phase) of a transformed previous iteration. Every iteration is converted to the base domain, and predefined known marker and padding data is reapplied to the iteration during every iteration, for convergence purposes. The restored image can be scaled and trimmed, and the marker position interpolated. The original image to be restored is modified to contain the known marker and known padding at known locations. These may be inserted in the base domain or in the transform domain.
Claims106
A system for processing an original image, comprising:
- a data modifier for inserting known data at one or more known locations of the image data in a base domain;
- a channel for communicating the original image as a certain component which constitutes less than all of multiple components in a transform domain;
- a storage for holding a plurality of iterated base-domain restored images, including an initial base-domain restored image containing the known data at the known location(s);
- a restore module for producing the iterated restored images by forming a combination of the certain component of the communicated image with a complementary component of a transform of the initial and previous ones of the iterated images, and by reapplying the known marker data to an inverse transform of the combination.
A system for processing an original image in a base domain, comprising:
- a data modifier for inserting known marker data at one or more known marker locations of the image in a base domain;
- a transform module for transforming the original image to less than all of a set of complementary components in a transform domain.
A system for processing an original image in a transform domain, comprising:
- a sensor for producing only one component of multiple complementary components in a transform domain;
- a data modifier for inserting known marker data at one or more known marker locations of the image in a base domain.
A system for processing a communicated image represented as less than all components of a transform domain, comprising:
- a storage for holding an initial restored image in a base domain, the initial restored image containing known marker data at one or more known marker location(s), and for holding a plurality of iterated base-domain restored images;
- a restore module for producing the iterated restored images by forming combinations of one component of the transform-domain communicated image with a complementary transform-domain component of the initial and previous ones of the iterated images, and by repeatedly reapplying the known marker data to inverse transforms of the combinations.
A method for processing an image, comprising:
- marking an original image so as to contain known marker data at one or more known marker locations of the image in a base domain;
- communicating the image as a certain component which constitutes less than all of multiple complementary components in a transform domain;
- generating an initial restored image in the base domain partly from the known marker data and location(s);
- for each of a determined number of iterations, with the initial restored image as the first of a series of restored images,
- transforming the restored image to the transform domain.
- combining the certain component of the communicated image with a complementary component of the transform-domain restored image;
- inverse transforming the restored image to the base domain;
- reapplying the known marker data at the known location(s) in the base-domain restored image.
A method for processing an original image in a base domain for subsequent restoration, comprising:
- selecting a known marker value related to predetermined data in the original image;
- placing the marker value at a known location with respect to the original image;
- transforming the original image to a transform domain having multiple complementary components;
- communicating less than all of the components as a modified image.
59where the padding data comprise zeros.
A method for processing an original image in a transform domain having multiple components for subsequent restoration, comprising:
- selecting a known marker value in a base domain related to predetermined data for the original image in the base domain;
- manipulating the original image entirely within the transform domain so as to place data therein representing a transform of the known base-domain marker value at a known location in the base domain;
- communicating less than all of the components of the manipulated image as a modified image.
A method for restoring an image to a base domain, comprising:
- receiving a modified image as one component of multiple complementary components in a transform domain, the one component containing predetermined known marker data at one or more marker locations in the base domain;
- generating an initial restored image in the base domain partly from the known marker data and location(s);
- for each of a number of iterations, with the initial restored image as the first of a series of restored images,
- transforming the restored image to the transform domain.
- combining the certain component of the communicated image with a complementary component of the transform-domain restored image;
- inverse transforming the restored image to the base domain;
- reapplying the known marker data at the known location(s) in the base-domain restored image;
- outputting one of the series of restored images as a final image in the base domain.
Dependencies are inferred from the claim text.