Spectral Resolution Enhancement of Hyperspectral Images via Sparse Representations

Fotiadou, Konstantina; Tsagkatakis, Grigorios; Tsakalides, Panagiotis

doi:10.2352/issn.2470-1173.2016.19.coimg-169

Cited by 11 publications

(16 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…A disruptive new concept in state-of-the-art signal processing refers to the sparse representations framework, where one seeks a sparsity-promoting decomposition of the input by exploiting available training observations. Particular applications of the sparse representations framework in spectral imaging involve denoising [17] and super-resolution [18], [19] among others. Recently, Degraux et al presented a 3D inpainting method for demosaicing SSI measurements, where reconstruction is achieved by exploiting appropriate sparsifying dictionaries, wavelets for the spatial dimension and DCT for the spectral [20].…”

Section: State-of-the-art Spectral Demosaicingmentioning

confidence: 99%

Graph and Rank Regularized Matrix Recovery for Snapshot Spectral Image Demosaicing

Tsagkatakis

Bloemen

Geelen

et al. 2019

IEEE Trans. Comput. Imaging

Self Cite

View full text Add to dashboard Cite

Snapshot Spectral Imaging (SSI) is a cutting-edge technology for enabling the efficient acquisition of the spatiospectral content of dynamic scenes using miniaturized platforms. To achieve this goal, SSI architectures associate each spatial pixel with a specific spectral band, thus introducing a critical trade-off between spatial and spectral resolution. In this paper, we propose a computational approach for the recovery of high spatial and spectral resolution content from a single or a small number of exposures. We formulate the problem in a novel framework of spectral measurement matrix completion and we develop an efficient low-rank and graph regularized method for SSI demosaicing. Furthermore, we extend state-of-the-art approaches by considering more realistic sampling paradigms that incorporate information related to the spectral profile associated with each pixel. In addition to reconstruction quality, we also investigate the impact of recovery on subsequent analysis tasks like classification using state-of-the-art convolutional neural networks. We experimentally validate the merits of the proposed recovery scheme using synthetically generated data from indoor and satellite observations and real data obtained with an IMEC visible range SSI camera.

show abstract

Section: State-of-the-art Spectral Demosaicingmentioning

confidence: 99%

Graph and Rank Regularized Matrix Recovery for Snapshot Spectral Image Demosaicing

Tsagkatakis

Bloemen

Geelen

et al. 2019

IEEE Trans. Comput. Imaging

Self Cite

View full text Add to dashboard Cite

show abstract

“…Redundancy Reduction: Because the dimension of HSI data is often large (tens of thousands), this causes a high computational cost in object detection applications, especially in deep learning approaches [14]. On the other hand, the sparseness of HSI data has been demonstrated in earlier works [15,16,20]. Therefore, selecting an appropriate subset of bands was considered an efficient process.…”

Section: Challenges and Pre-processing Techniquesmentioning

confidence: 99%

“…For example, among attempts using support vector machine (SVM) [32,2,46], a more recent work [36] was based on independent component analysis and morphological features. Meanwhile, sparsity-based algorithms [15,16,20] showed that the sparse representation of a pixel can predict the class label of the test sample better than classical SVMs. Recently, deeplearning approaches [14,40,17,53,28,35,39,41,51,52,50,43,48] make use of hierarchically extracted deep features.…”

Section: Introductionmentioning

confidence: 99%

Airborne Object Detection Using Hyperspectral Imaging: Deep Learning Review

Pham

Takalkar

et al. 2019

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Hyperspectral images have been increasingly important in object detection applications especially in remote sensing scenarios. Machine learning algorithms have become emerging tools for hyperspectral image analysis. The high dimensionality of hyperspectral images and the availability of simulated spectral sample libraries make deep learning an appealing approach. This report reviews recent data processing and object detection methods in the area including hand-crafted and automated feature extraction based on deep learning neural networks. We compare the accuracy performance based on existing reports as well as our own experiments. We found that CNN models provided reliable performance of over 97% detection accuracy across a large set of HSI collections. We included a wide range of data from a rural area (Indian Pines data), an urban area (Pavia University), a wetland region (Botswana), an industrial field (Kennedy Space Center), to a farm site (Salinas). Note that, the Botswana set was not reviewed in recent works, thus high accuracy selected methods were newly compared in this work. We also found that a plain CNN model can perform comparably to its more complex variants in target detection applications.

show abstract

“…Coupled dictionary learning relies on generating a pair of dictionaries which jointly encode the active S α , and the passive S p feature spaces, where the signals have sparse representations in terms of the trained dictionaries [16]. The main task is to recover a coupled dictionary pair D α and D p for the spaces S α and S p , respectively [10], and their corresponding sparse codes W α and W p , under the constraint W α = W p , by solving the following sparse matrix decomposition problems: argmin…”

Section: Coupled Dictionary Learningmentioning

confidence: 99%

Recovery of Soil Moisture Active Passive (SMAP) Instrument's Active Measurements via Coupled Dictionary Learning

Fotiadou¹,

Tsagkatakis²,

Moghaddam³

et al. 2018

Self Cite

View full text Add to dashboard Cite

NASA's Soil Moisture Active Passive (SMAP) satellite mission combines an L-band radiometer (passive) and a Synthetic Aperture Radar (SAR active) instrument in order to monitor the near-surface soil moisture and freeze-thaw states globally, with a revisit frequency of 23 days. Specifically, SMAP provides three soil moisture products, namely: one high-resolution from the radar, one low-resolution from the radiometer, and an intermediate-resolution from the fusion of the radar and radiometer. Unfortunately, SMAP's SAR instrument has halted its transmissions after a short operating period. In order to overcome this limitation, we introduce a novel post-acquisition computational technique aiming to synthesize the active measurement of SMAP's instrument, by exploiting the mathematical frameworks of Sparse Representations and Dictionary Learning. We propose a coupled dictionary learning model which considers joint feature spaces, composed of active and passive images, in order to recover the missing active measurements. We formulate our coupled dictionary learning problem within the context of the Alternating Direction Method of Multipliers. Experimental results demonstrate the ability of the proposed approach to reconstruct the active measurements, achieving better performance compared to state-of-the-art coupled dictionary learning techniques.

show abstract

Spectral Resolution Enhancement of Hyperspectral Images via Sparse Representations

Cited by 11 publications

References 24 publications

Graph and Rank Regularized Matrix Recovery for Snapshot Spectral Image Demosaicing

Graph and Rank Regularized Matrix Recovery for Snapshot Spectral Image Demosaicing

Airborne Object Detection Using Hyperspectral Imaging: Deep Learning Review

Recovery of Soil Moisture Active Passive (SMAP) Instrument's Active Measurements via Coupled Dictionary Learning

Contact Info

Product

Resources

About