Efficacy of Filtering Techniques in Improving Landsat SLC-Off Thermal Infrared Data

In 2003, Landsat-7 Enhanced Thematic Mapper Plus encountered a defect in its Scan Line Corrector (SLC) mechanism resulting in the loss of about 22% pixels in each image. Thermal Infrared band is used in applications such as temperature estimation and requires a single source image for estimation of dead pixels. This analysis tests the efficacy of various Radar and Morphological filters in reconstruction of dead pixels and improving the quality of Landsat SLC-off image. Efficacy has been based on visual analysis, histogram, and image quality metrics, namely, histogram error, root mean square error, peak signal to noise ratio, and Structural SIMilarity. Of all the experimented filters, nonlinear filters-Median, Local Region, and Morphological perform better. However, Morphological filter is able to eliminate dead pixels using a 7 * 7 kernel window while Median filter is able to do so with a 29 * 29 window. All four functions-Open, Close, Erode, and Dilate of Morphological filter are able to suppress noise (fill dead pixel gaps) while preserving edges and details resulting in sharp output images. On the basis of image quality metrics, use of Open Morphological filter is suggested.