Search
Menu
Hamamatsu Corp. - Earth Innovations LB 2/24

Software Fills the Blanks in Imagery

Facebook X LinkedIn Email
Daniel S. Burgess

As satellite-borne sensors have become more sophisticated, a wealth of data has become available to Earth scientists. Unfortunately, the level of detail these sensors afford carries a downside: It has become much more difficult and computationally intensive to estimate what had been in the occasional gaps that local conditions and technical problems inevitably produce, as a day’s data has increased to the level of gigabytes.

Software Fills the Blanks in Imagery
The dynamic multiresolution spatial model fills in the gap in ozone data by “zooming out” to generate estimates for the missing pixels at low resolution and then “zooming in” to refine the values. Courtesy of Ohio State University.

A new multiresolution spatial model overcomes these problems by appealing to the values of adjacent pixels collected at the same time that the gap in the data appeared as well as at previous times. A change-of-resolution Kalman filter establishes the optimal relative contribution from those pixels based on their reliability.

The algorithm and software were developed at The Ohio State University in Columbus by a team led by Noel Cressie, a professor of statistics and director of the Program in Spatial Statistics and Environmental Sciences, with his students Gardar Johannesson and Hsin-Cheng Huang. The latter two are now at Lawrence Livermore National Laboratory in California and at Academia Sinica in Taipei, Taiwan, respectively.

To ease the task of data processing, the software first generates potential values for the blank pixels in an image at a low resolution. Based on these values, it produces estimates for the pixels at higher resolutions.

In tests of the technique using spectroscopic measurements of total column ozone from October 1988 taken with an instrument on the Nimbus-7 satellite, the researchers found that it filled a gap in one day’s data (approximately 160,000 observations) in about three minutes. They employed a single-processor workstation in the demonstration. They also used the software to estimate the missing data in the measurements from the entire month.

A report of the work will appear in a future issue of Environmental and Ecological Statistics.
Cognex Corp. - Smart Sensor 3-24 GIF MR

Published: January 2006
multiresolution spatial modelpixelsResearch & Technologysatellite-borne sensorsSensors & DetectorsTech Pulse

We use cookies to improve user experience and analyze our website traffic as stated in our Privacy Policy. By using this website, you agree to the use of cookies unless you have disabled them.