Removal of Thin Clouds Using Cirrus and QA Bands of Landsat-8

Removal of Thin Clouds in Landsat-8 OLI Data with Independent Component Analysis

“…In addition, a coastal/aerosol band (Band 1, 433-453 nm of wavelength), a cirrus band (Band 9, 1360-1390 nm of wavelength), and a quality assessment (QA) band are added to Landsat-8 so that the needs of scientific studies and routine operations can be better met. The inclusion of Band 9 and QA band also stimulates the study of the cloud removal using one single Landsat-8 image (e.g., [2,3]). …”

“…Thus, the satellite remote sensing data in the optical region are affected. Without the correction of the cloud effect, results derived from the data can be erroneous (e.g., [3,4]). Therefore, the removal of cloud effect is necessary.…”

“…The filter could affect the spectral characteristics of pixels in cloud-free areas as well. The image replacement approach uses either multi-temporal images [7][8][9] or a single image [3,10]. As long as there is a cloud-free image available on another acquisition date, the multi-temporal image replacement is feasible.…”

“…Coupled with Band 9 and QA band, Shen et al (2015) [3] developed an image replacement approach using one single image to remove clouds for Bands 1-5 of Landsat-8. However, there are four main concerns with that method [3].…”

“…However, there are four main concerns with that method [3]. First, QA band data are used, and thermal infrared (TIR) data are needed to create the QA band.…”