Bayesian Estimation of High-Resolution Imagery from Low-Resolution Video Sequences and Multisensor Data Sets
Abstract
In modern spy movies such as Enemy of the State and Patriot Games, scientists magnify digital satellite imagery to determine the identities of people captured on surveillance video. These individuals appear rather blocky at the coarsest resolution scales because of undersampling by the image sensor array, but they become remarkably clear after zooming in on a particular region-of- interest. Although real-world video enhancement algorithms are not capable of calculating the perfect results produced in Hollywood, additional details can be extracted from an image sequence by integrating several neighboring frames that contain subpixel-resolution scene/object motion. The five most important research results obtained through this grant activity related to the Bayesian estimation of high-resolution imagery from low-resolution digital video include: (1) nonlinear filtering of subpixel motion vectors for improved super-resolution video enhancement; (2) estimation of subpixel motion fields from segmented image sequences; (3) super-resolution enhancement of compressed digital video; (4) multiframe integration via the projective transformation with automated block matching feature point pair selection; and (5) development of a Windows-based video enhancement software tool. As a follow-up to these R&D activities, the PI has been attempting to secure funding to market super- resolution enhancement technologies to law enforcement and national security agencies.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jan 08, 2002
- Accession Number
- ADA398302
Entities
People
- Richard R. Schultz
Organizations
- University of North Dakota