Multiresolutional Optic Flow
Abstract
The irony embedded in optic flow computation is that despite its not having been studied mathematically until about 20 years ago, humans have been performing the related calculations innately since their creation. The process can be thought of as filling in the gaps that occur between the images we see changing in time. For the human mind, the transition is automatic from what an image looks like at one instant to what it looks like at the next. Our brain intuitively generates a continuous flow of what it perceives is being viewed by the eyes. Manifested in a mathematical sense, optic flow serves as a mechanism to describe the movement of objects in a digital image sequence using a flow field of grayscale intensity. This particular study has focused upon developing a usable mathematical implementation for two of the various algorithms, which have been proposed to compute optic flow. One method relies upon calculus of variations to regularize the problem while the other utilizes a wavelet based multi-scale approach. Possible applications for such research range from data compression of video sequences to the development of a more efficient way to analyze the time varying one-dimensional acoustic imagery supplied by sonar systems.
Document Details
- Document Type
- Technical Report
- Publication Date
- May 02, 2003
- Accession Number
- ADA416020
Entities
People
- Matthew J. Ahlert
Organizations
- United States Naval Academy