Image Sequence Enhancements Based on the Normal Component of Image Motion
Abstract
In a recent paper, Huang and Hsu describe an enhancement algorithm for time-varying images. This algorithm achieves the type of enhancement that one would obtain from averaging multiple images of a static scene by using motion vectors computed from the sequence to establish the correspondence between pixels in successive frames. The purpose of this note is to demonstrate how and why it is often sufficient to consider only the normal components of the motion vectors, i.e., the component in the gradient direction, to compute such enhancements. There is a computational advantage to using only the normal component, since it is directly measurable from the image sequence by combining the space and time derivatives of image intensity, while computing the actual motion vectors ordinarily requires some spatial integration of the normal components or the application of a local correlation procedure. Furthermore, in order to compute the motion vectors one must make some assumption about the local image motion (for example, that the image motion is a two-dimensional translation); when this assumption is incorrect, the computed motion vectors will be in error.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jul 01, 1982
- Accession Number
- ADA128652
Entities
People
- Azriel Rosenfeld
- Hu-chen Xie
- Larry S. Davis
Organizations
- University of Maryland