Algorithm Development for Full Gaps of Landsat 7 Images

Abstract

Landsat7 of Enhanced thematic mapper plus (ETM+) was launched on April 15, 1999. Four years later, images start degrading due to the scan line corrector (SLC). SLC is a malfunction that results in pixel gaps in images captured by the sensor of Landsat7. The pixel gap regions extend from about one pixel near the image center and reach up to about 14 pixels in width near the image edge. The shape of this loss is like a zigzag line; however, there are different studies about repairing these gaps. The challenge of all studies depends on retrieving inhomogeneous areas because the homogenous area can be retrieved quickly depending on the surrounding area. This research focuses on filling these gaps by utilizing pixels around them gaps. Pixels from different satellites with the same resolution at the approximate equal date and exact location would be utilized. Mean and standard deviation were used in gap pixel predicting. For complete pixels gap predicting, an image registration should be applied for predicted a good result—image registration comprised of geometric and radiometric registration. Translation and rotation was the main effect of geometric registration, while pixel tone was mainly related to radiometric registration. Two images of Landsat8 (before and after) with nearly the date of Landsat 7 were used. The fidelity criteria (mean square error and mean absolute error) were utilized to evaluate the results. The results showed a good prediction of the gap pixels in the Landsat 7 image, considering the geometric and radiometric parts.