Skip to main content
ARS Home » Research » Publications at this Location » Publication #273613
Title: An enhanced neighborhood similar pixel interpolator approach for removing thick clouds in landsat images

Author
item ZHU, XIAOLIN - The Ohio State University
item Gao, Feng
item LIU, DESHENG - The Ohio State University
item CHEN, JIN - Beijing Normal University

Submitted to: Geoscience and Remote Sensing Letters
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 3/15/2012
Publication Date: 5/15/2012
Citation: Zhu, X., Gao, F.N., Liu, D., Chen, J. 2012. An enhanced neighborhood similar pixel interpolator approach for removing thick clouds in landsat images. Geoscience and Remote Sensing Letters. 9(3):521-525.

Interpretive Summary: Land remote sensing is often limited by the thick clouds that block thr land surface from the satellite observation. In order to make a cloud-free image, several images from different acquisition dates are normally used to composite a clear image. The traditional compositing approaches require images come from a very short period of time by assuming that the land surface has not changed. However, this assumption is not realistic for Landsat data since Landsat revisits the same location on a 16-day cycle. In this paper, we developed a new improved approach to replace the cloudy portion of Landsat image from previous clear observations. The approach has been tested on both the simulated and real Landsat data. Results show that the approach can restore the cloudy portion of the image well. This work provides a feasible way to reconstruct a clear Landsat image from a partially cloud contaminated image and thus make it more valuable for a more frequent land surface monitoring such as for crop condition monitoring.

Technical Abstract: Thick cloud contaminations in Landsat images limit their regular usage for land applications. A few methods have been developed to remove thick clouds using additional cloud-free images. Unfortunately, the cloud-free composition image produced by existing methods commonly lacks from the desired spatial continuity. Based on the assumption that the neighboring pixels outside cloudy patches have similar temporal change patterns to the cloudy pixels if their spectral characteristics are similar, this paper presents an improved neighborhood similar pixel interpolator (NSPI) approach to build cloud-free imagery. NSPI approach was originally developed and tested for filling gaps due to the Landsat ETM+ Scan Line Corrector (SLC)-off problem. Both the simulated and real cloudy images were used to evaluate the performance of the proposed method. The results show that NSPI approach can restore the reflectance of cloud contaminated images with fewer artifact edge effects compared to a contextual multiple linear prediction (CMLP) method. The reflectance restored by NSPI approach is more effective especially when the cloud-free auxiliary image and cloudy image are acquired from different seasons and have different spectral characteristics.