Patch-Based Information Reconstruction of Cloud-Contaminated Multitemporal Images

Lin, Cheng-Yao; Lai, Kang-Hua; Chen, Zhibin; Chen, Jyun-Yuan

doi:10.1109/tgrs.2012.2237408

Cited by 93 publications

(46 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Substitution approaches were used to fill in cloudy pixels using information from adjacent cloud-free pixels in the same image. Otherwise information was retrieved for the corresponding pixels from previous time periods [25,26]. In addition to methods applied directly to NDVI, data recovery techniques for other types of similar data are also informative.…”

Section: Data Recovery Using Empirical Methodsmentioning

confidence: 99%

“…Multi-temporal dictionary learning algorithms have been able to efficiently reduce clouds and accurately reconstruct contaminated surficial information underlying large-scale clouds and shadows [29]. A new patch matching multi-temporal group sparse representation (PM-MTGSR) method has also been proposed for the reconstruction of the optical remote sensing missing data [26]. They suggested that the method is suitable for thick cloud cover or cases of sensor failure.…”

Section: Data Recovery Using Empirical Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Optical Cloud Pixel Recovery via Machine Learning

et al. 2017

View full text Add to dashboard Cite

Remote sensing derived Normalized Difference Vegetation Index (NDVI) is a widely used index to monitor vegetation and land use change. NDVI can be retrieved from publicly available data repositories of optical sensors such as Landsat, Moderate Resolution Imaging Spectro-radiometer (MODIS) and several commercial satellites. Studies that are heavily dependent on optical sensors are subject to data loss due to cloud coverage. Specifically, cloud contamination is a hindrance to long-term environmental assessment when using information from satellite imagery retrieved from visible and infrared spectral ranges. Landsat has an ongoing high-resolution NDVI record starting from 1984. Unfortunately, this long time series NDVI data suffers from the cloud contamination issue. Though both simple and complex computational methods for data interpolation have been applied to recover cloudy data, all the techniques have limitations. In this paper, a novel Optical Cloud Pixel Recovery (OCPR) method is proposed to repair cloudy pixels from the time-space-spectrum continuum using a Random Forest (RF) trained and tested with multi-parameter hydrologic data. The RF-based OCPR model is compared with a linear regression model to demonstrate the capability of OCPR. A case study in Apalachicola Bay is presented to evaluate the performance of OCPR to repair cloudy NDVI reflectance. The RF-based OCPR method achieves a root mean squared error of 0.016 between predicted and observed NDVI reflectance values. The linear regression model achieves a root mean squared error of 0.126. Our findings suggest that the RF-based OCPR method is effective to repair cloudy pixels and provides continuous and quantitatively reliable imagery for long-term environmental analysis.

show abstract

Section: Data Recovery Using Empirical Methodsmentioning

confidence: 99%

Section: Data Recovery Using Empirical Methodsmentioning

confidence: 99%

Optical Cloud Pixel Recovery via Machine Learning

et al. 2017

View full text Add to dashboard Cite

show abstract

“…Of course, many approaches have been designed to overcome this problem (vanDijk et al, 1987;Viovy et al, 1992;Roerink et al, 2000;Chen et al, 2004;Beck et al, 2006;Ma and Veroustraete, 2006;Hird and McDermid, 2009;Julien and Sobrino, 2010;Cho and Suh, 2013;Ke et al, 2013;Lin et al, 2014;Michishita et al, 2014;Moreno et al, 2014;Xiao et al, 2015;Xu et al, 2015;Yang et al, 2015;Zhou et al, 2015). These approaches, termed time series reconstruction -or gap-filling-methods, aim at providing alternative data for cloud-contaminated observations.…”

Section: Introductionmentioning

confidence: 99%

TISSBERT: A benchmark for the validation and comparison of NDVI time series reconstruction methods

Julien

Sobrino

2018

Rev. Teledetec.

View full text Add to dashboard Cite

This paper introduces the Time Series Simulation for Benchmarking of Reconstruction Techniques (TISSBERT) dataset, intended to provide a benchmark for the validation and comparison of time series reconstruction methods. Such methods are routinely used to estimate vegetation characteristics from optical remotely sensed data, where the presence of clouds decreases the usefulness of the data. As for their validation, these methods have been compared with previously published ones, although with different approaches, which sometimes lead to contradictory results. We designed the TISSBERT dataset to be generic so that it could simulate realistic reference and cloud-contaminated time series at global scale. To that end, we estimated both cloud-free and cloud-contaminated Normalized Difference Vegetation Index (NDVI) statistics for randomly selected control points and each day of the year from the Long Term Data Record Version 4 (LTDR-V4) dataset by assuming different statistical distributions. The best approach was then applied to the whole dataset, and validity of the results were estimated through the Kolmogorov-Smirnov statistic. The dataset elaboration is described thoroughly along with how to use it. The advantages and drawbacks of this dataset are then discussed, which emphasize the realistic simulation of the cloud-contaminated and reference time series. This dataset can be obtained from the authors upon demand. It will be used in a next paper to compare widely used NDVI time series reconstruction methods.Key words: NDVI, gap-filling, reconstruction, dataset, comparison.TISSBERT: una referencia para la validación y la comparación de métodos para la reconstrucción de series temporales de NDVI Resumen: En este trabajo se presenta la base de datos titulada Time Series Simulation for Benchmarking of Reconstruction Techniques (TISSBERT) con el propósito de ofrecer una herramienta para la validación y la comparación de métodos para la reconstrucción de series temporales. Tales métodos se usan de manera rutinaria para la estimación de características de la vegetación a partir de datos obtenidos por teledetección óptica, donde la presencia de nubes disminuye su utilidad. En cuanto a su validación, estos métodos se han comparado con otros publicados anteriormente, aunque desde perspectivas diferentes, lo cual conduce a resultados contradictorios. La base de datos TISSBERT se ha diseñado como una herramienta genérica para una simulación realista a escala global de series temporales de referencia o contaminadas por nubes. Para ello, se estimaron estadísticas de Normalized Difference Vegetation Index (NDVI) con y sin contaminación de nubes para unos píxeles de control seleccionados de manera aleatoria, y para cada día del año, usando la base de datos Long Term Data Record Version 4 (LTDR-V4), y probando con varias distribuciones estadísticas. La mejor metodología se aplicó al conjunto de la base de datos, y la validez de

show abstract

“…These existing methods can be divided into three categories: (1) spatial-based methods; (2) spectralbased methods; and (3) temporal-based methods [1].…”

Section: Introductionmentioning

confidence: 99%

“…Lin et al proposed replacing the missing image using optimal information from the same region in a referenced image when the time interval between the missing image and the referenced image is short enough [1]. The method appeared to demonstrate better results than a method proposed by the joint U.S. Geological Survey (USGS) Landsat team to reconstruct data gaps in SLC-off images (i.e., images taken when the Scan Line Corrector (SLC) is not available) [8]; however, the result of this method depends greatly on the selected sampling, and therefore its accuracy is uncertain.…”

Section: Introductionmentioning

confidence: 99%

Reconstruction of MODIS Spectral Reflectance under Cloudy-Sky Condition

et al. 2016

View full text Add to dashboard Cite

Clouds usually cause invalid observations for sensors aboard satellites, which corrupts the spatio-temporal continuity of land surface parameters retrieved from remote sensing data (e.g., MODerate Resolution Imaging Spectroradiometer (MODIS) data) and prevents the fusing of multi-source remote sensing data in the field of quantitative remote sensing. Based on the requirements of spatio-temporal continuity and the necessity of methods to restore bad pixels, primarily resulting from image processing, this study developed a novel method to derive the spectral reflectance for MODIS band of cloudy pixels in the visual-near infrared (VIS-NIR) spectral channel based on the Bidirectional Reflectance Distribution Function (BRDF) and multi-spatio-temporal observations. The proposed method first constructs the spatial distribution of land surface reflectance based on the corresponding BRDF and the solar-viewing geometry; then, a geographically weighted regression (GWR) is introduced to individually derive the spectral surface reflectance for MODIS band of cloudy pixels. A validation of the proposed method shows that a total root-mean-square error (RMSE) of less than 6% and a total R 2 of more than 90% are detected, which indicates considerably better precision than those exhibited by other existing methods. Further validation of the retrieved white-sky albedo based on the spectral reflectance for MODIS band of cloudy pixels confirms an RMSE of 3.6% and a bias of 2.2%, demonstrating very high accuracy of the proposed method.

show abstract

Patch-Based Information Reconstruction of Cloud-Contaminated Multitemporal Images

Cited by 93 publications

References 18 publications

Optical Cloud Pixel Recovery via Machine Learning

Optical Cloud Pixel Recovery via Machine Learning

TISSBERT: A benchmark for the validation and comparison of NDVI time series reconstruction methods

Reconstruction of MODIS Spectral Reflectance under Cloudy-Sky Condition

Contact Info

Product

Resources

About