Nonlinear Pixel Replacement Estimation.
Abstract
Optical surveillance sensors are sometimes subject to static noise processes which can adversely affect the optimal processing of their resulting imagery. For example, image intensity values derived from line outages, dead pixels, popcorn noise and other such noise mechanisms will contaminate both local and global estimates of the power spectrum density, probability density function and other key statistical properties inherent to the original observed scene. If these estimates are used to derive optimum filters for detecting specific targets in said imagery or registering sequences of images, poor processing performance could result. References 1 through 7 provide excellent reviews of current image processing trends dependent on good quality pictures and illustrate their utility for enhancing the inherent information content found in remotely sensed images such as those taken by the LANDSAT and NIMBUS-7 satellites. They also show the effect of noisy pixels on these techniques and the types of performance degradations incurred; which can be significant. This suggests that methods for replacing the bad pixel values with numbers commensurate with the inherent statistics of a detected image can be important to optimum image processing in many applications.
Document Details
- Document Type
- Technical Report
- Publication Date
- Apr 01, 1986
- Accession Number
- ADA168882
Entities
People
- R. Cigledy