IMAGE SMOOTHING WITH DECOUPLED REGULARIZATION

Patent №

US 7,206,440

Granted

2007-04-17

Filed 2003

Owner

CARNEGIE MELLON UNIVERSITY

AI components

2

vision · kr

Assignment

Recorded

Dataset

AIPD

2023_r1 edition

Application

10366742

A technique for simplifying the processing of tomographic imagery (e.g., CT and PET) via statistical smoothing is disclosed. An image smoothness criterion is imposed in the projection domain while identifying data that matches observed data within a desired tolerance. The present invention thus amounts to a computation entirely in the projection domain. As a result, the standard (and somewhat expensive) practice of numerically iterating between the image and projection domains is avoided. Also, the fundamental system of equations having on the order of m p unknowns is decoupled into p independent systems of m unknowns each, where there are p projections of m measurements each. The present invention thus provides a processing technique defined in a single domain, which may be carried out via parallel processing.

VisionKnowledge representationG06T 11/005G06T 12/10Y10S 378/901

AI classification

Vision1.00
Knowledge representation0.59
AI hardware0.35
Planning0.04
Natural language0.01
Machine learning0.00
Speech0.00
Evolutionary computation0.00

Ownership

CARNEGIE MELLON UNIVERSITY

assignment · 142130986

Assignors

AUGUST, JONAS

On an employer assignment, the assignors are typically the inventors.

© 2026 NYSGPT2525 LLC