Cloud Removal in Image Time Series Through Sparse Reconstruction From Random Measurements

Cerra, Daniele; Bieniarz, Jakub; Beyer, Florian; Tian, Jiaojiao; Müller, Rupert; Jarmer, Thomas; Reinartz, Peter

doi:10.1109/jstars.2016.2550084

Cited by 11 publications

(10 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The three bands with a spatial resolution of 60 meters are discarded, and the bands at higher resolution are resampled to 20 meters after applying a Gaussian filter separately to each one to prevent aliasing. Please note that this is done only to simplify the workflow: the bands could be kept at their original resolution by applying the method separately to the three groups of spectral bands, with no degradations in performance (Cerra et al, 2016).…”

Section: Resultsmentioning

confidence: 99%

“…2. The number of elements |D| for each dictionary D is chosen applying the empirical rule defined in (Cerra et al, 2016): In which N is the dimensionality of the subspace containing the relevant information in the dataset. To estimate N methods traditionally applied to hyperspectral image processing could be used (Bioucas-Dias and Nascimento, 2008).…”

Section: Resultsmentioning

confidence: 99%

“…the setting of the λ regularization parameter, a value of 1 usually yields satisfactory results whenever the dictionary entries are collected from the same image (Cerra et al, 2016). With the dictionary size empirically set as will be detailed in next Section, the only parameter requiring further investigations is the number of iterations.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Cloud Removal From Sentinel-2 Image Time Series Through Sparse Reconstruction From Random Samples

Cerra¹,

Bieniarz²,

Müller³

et al. 2016

Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci.

View full text Add to dashboard Cite

ABSTRACT:In this paper we propose a cloud removal algorithm for scenes within a Sentinel-2 satellite image time series based on synthetisation of the affected areas via sparse reconstruction. For this purpose, a clouds and clouds shadow mask must be given. With respect to previous works, the process has an increased automation degree. Several dictionaries, on the basis of which the data are reconstructed, are selected randomly from cloud-free areas around the cloud, and for each pixel the dictionary yielding the smallest reconstruction error in non-corrupted images is chosen for the restoration. The values below a cloudy area are therefore estimated by observing the spectral evolution in time of the non-corrupted pixels around it. The proposed restoration algorithm is fast and efficient, requires minimal supervision and yield results with low overall radiometric and spectral distortions.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Resultsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Cloud Removal From Sentinel-2 Image Time Series Through Sparse Reconstruction From Random Samples

Cerra¹,

Bieniarz²,

Müller³

et al. 2016

Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci.

View full text Add to dashboard Cite

show abstract

“…Two multi-temporal dictionary learning algorithms have been expanded from the original dictionary learning to the recovery of cloud and shadow regions without manually designed parameters [29]. Cerra et al [30] introduced sparse representation theory into cloud removal and reconstructed the dictionary randomly from the available elements of the temporal image. Xu et al [31] proposed multi-temporal dictionary learning (MDL) to learn the cloudy areas (target data) and the cloud-free areas (reference data) separately in the spectral domain.…”

Section: Learning-based Cloud Removal Approachesmentioning

confidence: 99%

Gated Convolutional Networks for Cloud Removal From Bi-Temporal Remote Sensing Images

Dai

Zhang

2020

Remote Sensing

View full text Add to dashboard Cite

Pixels of clouds and cloud shadows in a remote sensing image impact image quality, image interpretation, and subsequent applications. In this paper, we propose a novel cloud removal method based on deep learning that automatically reconstructs the invalid pixels with the auxiliary information from multi-temporal images. Our method’s innovation lies in its feature extraction and loss functions, which reside in a novel gated convolutional network (GCN) instead of a series of common convolutions. It takes the current cloudy image, a recent cloudless image, and the mask of clouds as input, without any requirements of external training samples, to realize a self-training process with clean pixels in the bi-temporal images as natural training samples. In our feature extraction, gated convolutional layers, for the first time, are introduced to discriminate cloudy pixels from clean pixels, which make up for a common convolution layer’s lack of the ability to discriminate. Our multi-level constrained joint loss function, which consists of an image-level loss, a feature-level loss, and a total variation loss, can achieve local and global consistency both in shallow and deep levels of features. The total variation loss is introduced into the deep-learning-based cloud removal task for the first time to eliminate the color and texture discontinuity around cloud outlines needing repair. On the WHU cloud dataset with diverse land cover scenes and different imaging conditions, our experimental results demonstrated that our method consistently reconstructed the cloud and cloud shadow pixels in various remote sensing images and outperformed several mainstream deep-learning-based methods and a conventional method for every indicator by a large margin.

show abstract

“…The algorithm learns multitemporal dictionary using the cloud-free parts of the images. Instead of random dictionaries as in the case of [7], the dictionary for the cloudy image is calculated using iterative optimization method. Cheng et al [8] present an algorithm that fills the cloud contaminated areas pixel by pixel.…”

Section: Introductionmentioning

confidence: 99%

Cloud Removal in High Resolution Multispectral Satellite Imagery: Comparing Three Approaches

Adam

Mönks

Esch

et al. 2018

The 2nd International Electronic Conference on Remote Sensing

View full text Add to dashboard Cite

Clouds and cloud-shadow are a persistent problem in all optical satellite imagery. Plenty of methods have been suggested in the literature to address this problem, and reconstruct the missing part of the optical signal. In this work, three methods representative of different approaches to the cloud removal problem are compared. The first method is temporal fitting using Fourier series, which benefits from the temporal continuity of the signal. The second method uses sparse spectral unmixing to fill in the missing areas. The third method employs radiometric consistency as a tool to determine the missing part of the signal. These three methods are first presented and their theoretical background described, followed by a discussion of their implied assumptions and general performance. A set of experiments using Landsat 8 time series with diverse land cover types were conducted. The quantitative results of the three methods using simulated clouds are presented. Finally, some concluding remarks about the relative advantages of the three approaches are listed, in addition to some recommendations about their use.

show abstract

Cloud Removal in Image Time Series Through Sparse Reconstruction From Random Measurements

Cited by 11 publications

References 35 publications

Cloud Removal From Sentinel-2 Image Time Series Through Sparse Reconstruction From Random Samples

Cloud Removal From Sentinel-2 Image Time Series Through Sparse Reconstruction From Random Samples

Gated Convolutional Networks for Cloud Removal From Bi-Temporal Remote Sensing Images

Cloud Removal in High Resolution Multispectral Satellite Imagery: Comparing Three Approaches

Contact Info

Product

Resources

About