Study of Digital Matching of Dissimilar Images.
Abstract
In this report, the registration of digital images from similar and dissimilar sensors is studied. The matching problem is divided into four subproblems: (1) detecting and extracting appropriate features; (2) determining an approximate global registration; (3) determining disparity for a subset of points in the image; and (4) determining the global nonlinear transformation for the entire image. Three classes of features and their detectors are considered: edge features, point features, and region features. Several registration methods are described, including a number of correlation procedures, the L.N.K. registration technique and its 3-d extension, and some region matching procedures. The registration techniques are briefly compared and contrasted. Their ability to handle the full rotation, scale and translation transformations, and their time and space requirements are considered. It is concluded that only two registration techniques are worthy of further investigation: the L.N.K. registration technique and a combined hierarchical-sequential correlation method using edge images.
Document Details
- Document Type
- Technical Report
- Publication Date
- Nov 01, 1980
- Accession Number
- ADA102619
Entities
People
- Barbara A. Lambird
- David Lavine
- George C. Stockman
- Kenneth C. Hayes
- Laveen N. Kanal