Hyperspectral image compression approaches: opportunities, challenges, and future directions: discussion

Rui Dusselaar; Manoranjan Paul

doi:10.1364/JOSAA.34.002170

Journal of the Optical Society of America A
Vol. 34,
Issue 12,
pp. 2170-2180
(2017)
•https://doi.org/10.1364/JOSAA.34.002170

Hyperspectral image compression approaches: opportunities, challenges, and future directions: discussion

Rui Dusselaar and Manoranjan Paul

Not Accessible

Your library or personal account may give you access

Get PDF
Email
Share
Get Citation
Copy Citation Text
Rui Dusselaar and Manoranjan Paul, "Hyperspectral image compression approaches: opportunities, challenges, and future directions: discussion," J. Opt. Soc. Am. A 34, 2170-2180 (2017)

Export Citation
- BibTex
- Endnote (RIS)
- HTML
- Plain Text
Citation alert
Save article

Check for updates

Abstract

This paper establishes a review of the recent study in the field of hyperspectral (HS) image compression approaches. Recently, image compression techniques have achieved significant advances from diverse types of coding standards/approaches. HS image compression requires an unconventional coding technique because of its unique, multiple-dimensional structure. The data redundancy exists in both inter-band and intra-band methods. The survey summarizes current literature in inter- and intra-band compression methods. The challenges, opportunities, and future research possibilities regarding HS image compression are further discussed. The experimental results are also provided for validity and applicability of the existing HS image compression techniques.

Full Article | PDF Article

More Like This

Built-up index methods and their applications for urban extraction from Sentinel 2A satellite data: discussion

Juan C. Valdiviezo-N, Alejandro Téllez-Quiñones, Adan Salazar-Garibay, and Alejandra A. López-Caloca
J. Opt. Soc. Am. A 35(1) 35-44 (2018)

CP tensor-based compression of hyperspectral images

Leyuan Fang, Nanjun He, and Hui Lin
J. Opt. Soc. Am. A 34(2) 252-258 (2017)

Clarification and unification of the obliquity factor in diffraction and scattering theories: discussion

Yijun Bao and Thomas K. Gaylord
J. Opt. Soc. Am. A 34(10) 1738-1745 (2017)

Previous Article Next Article

Cited By

You do not have subscription access to this journal. Cited by links are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Figures (4)

You do not have subscription access to this journal. Figure files are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Tables (3)

You do not have subscription access to this journal. Article tables are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Equations (1)

You do not have subscription access to this journal. Equations are available to subscribers only. You may subscribe either as an Optica member, or as an authorized user of your institution.

Contact your librarian or system administrator
or
Login to access Optica Member Subscription

Algorithms	Authors	Contribution
Inter-band compression
KLT/PCA transform	Du and Fowler [21] and Wang et al. [22]	PCA-based compression techniques discard part of the image signal. This may not be suitable for specific applications. Also, one of the major difficulties in applying PCA is determining the optimal number of PCs for implementation.
Pairwise orthogonal transform	Nian et al. [19]	It overcomes the problem of KLT, such as bit depth expansion and lack of scalability, and reduces memory requirements.
Band-selection-based inter-bands compression	Wang et al. [75], Zhu et al. [15], and Zhao et al. [14]	Selecting an efficient GOB measure is still a vital choice. There is no universal metric of GOBs applicable to all HS images at different wavelengths. Moreover, some of the experimental results [20,21] show that the algorithms can only achieve effective compression at low bit rates.
Tensor	Zhang et al. [8] and Veganzones et al. [9]	Tensors can maintain the spatial structure of data. However, a huge amount of spectral information contained in an HS image to generate a tensor by using a multiplicative update algorithm makes computation extremely complicated and sometimes even unfeasible. To reduce the computational cost and speed up the convergence is very important.
Intra-band compression
Vector quantization transform	Zhang et al. [33], Zhao and Jing [34], and Li et al. [35]	VQ can achieve a compression result by reducing color distribution. However, the design of a codebook is considered to be a challenging problem.
Discrete cosine transform/3DCT	Rasti et al. [36] and Qiao et al. [17]	One of the traditional compression techniques. It can remove intra-band correlation, but inter-band correlation is left untreated. Therefore, the 3D-DCT approach is introduced. It is an enhancement of DCT to extend the multiple-dimensional process level. However, the reconstructed image undergoes blocking effects, which severely degrade visual quality.
3D wavelets transforms	Tang and Pearlman [42], Shingate et al. [39], Zala and Parmar [41], and George and Manimekalai [40]	3D transform is not specifically designed for features of an HS image. It does not fully utilize the common background of inner bands and spectral redundancy.
Dictionary learning and prediction-based compression	Lin and Hwang [59], Castrodad et al. [56], Lin and Hwang [58], and Ülkü and Töreyin [55]	Dictionary-based learning methods take consideration of inherent characteristics of the input data. Uses only a few number of dictionary elements to present the dataset. It is a data-driving method. Using a dictionary-generating process is a key step to determining the compression performance.
Video coding H.264/HEVC	Gao et al. [62], Dusselaar et al. [63], and Podder et al. [76]	The HEVC standard has shown significant improvement over the state-of-the-art transformation-based still-image coding standard with spectral prediction modeling.

HS Library	Provider	Spectral Range	Description
AVIRIS—Airborne Visible Infrared Imaging Spectrometer	NASA	0.4–2.5 μm	AVIRIS is the most widely used HS database. AVIRIS data are basically related to the global environment and climate change; http://aviris.jpl.nasa.gov/.
HYDICE—Hyperspectral Digital Imagery Collection Experiment	Naval Research Laboratory	0.4–2.5 μm	HYDICE is a 206 channel imaging spectrometer. The data were collected on an ERIM CV 580 aircraft and a C-141 aircraft at altitudes up to 40,000 feet [78].
Natural scenes dataset	University of Manchester Foster, D. H., etc.	0.4–0.7 μm	The natural scenes dataset consists of a mixture of rural scenes from the Minho region of Portugal, containing rocks, trees, leaves, grass, and urban scenes from the cities of Porto and Braga; http://personalpages.manchester.ac.uk/staff/david.foster/Hyperspectral_images_of_natural_scenes_04.html.
Purdue’s Indianan Indian Pine test site	Purdue University Research Repository	0.4–2.5 μm	Purdue’s Indianan Indian Pine test site is another widely used HS image. It provides 12 band moderate dimension images and 220 band HS images; https://engineering.purdue.edu/~biehl/MultiSpec/hyperspectral.html.
Scene data	SCIEN, Stanford University	0.4–2.5 μm	The dataset contains images of faces, landscapes, and buildings. The images have been recorded with two HySpex line-scan imaging spectrometers, covering the spectral ranges of 0.4–1 μm and 1–2.5 μm; https://scien.stanford.edu/index.php/hyperspectral-image-data/.
USGS Digital Spectral Library	USGS Spectroscopy Laboratory	0.4–2.5 μm	USGS is a very useful laboratory data library, providing various remote sensing spectrometers; https://speclab.cr.usgs.gov/spectral-lib.html.

Images	Low Bit Rates 0.5 bpppb
Images	(35 PSNR or More)
	F. Zhao’s [61]	AT-3DSPIHT [47]	AT-3DSPECK [80]	EZW	ASWDR	3D-SPIHT
Cuprite_1	52.8	51.4	52.07	39.39	39.51	39.57
Jasper Ridge	52.76	51.16	51.11	36.47	38.65	39.85
Lunar Lake	52.22	51.74	52.01	33.56	29.87	31.24
High bit rate 1 bpppb
Cuprite_1	55.49	54.9	54.32	44.35	47.24	48.44
Jasper Ridge	56.34	56.08	55.93	43.27	45.56	48.57
Lunar Lake	54.71	54.27	55.01	43.45	44.74	49.69

Algorithms	Authors	Contribution
Inter-band compression
KLT/PCA transform	Du and Fowler [21] and Wang et al. [22]	PCA-based compression techniques discard part of the image signal. This may not be suitable for specific applications. Also, one of the major difficulties in applying PCA is determining the optimal number of PCs for implementation.
Pairwise orthogonal transform	Nian et al. [19]	It overcomes the problem of KLT, such as bit depth expansion and lack of scalability, and reduces memory requirements.
Band-selection-based inter-bands compression	Wang et al. [75], Zhu et al. [15], and Zhao et al. [14]	Selecting an efficient GOB measure is still a vital choice. There is no universal metric of GOBs applicable to all HS images at different wavelengths. Moreover, some of the experimental results [20,21] show that the algorithms can only achieve effective compression at low bit rates.
Tensor	Zhang et al. [8] and Veganzones et al. [9]	Tensors can maintain the spatial structure of data. However, a huge amount of spectral information contained in an HS image to generate a tensor by using a multiplicative update algorithm makes computation extremely complicated and sometimes even unfeasible. To reduce the computational cost and speed up the convergence is very important.
Intra-band compression
Vector quantization transform	Zhang et al. [33], Zhao and Jing [34], and Li et al. [35]	VQ can achieve a compression result by reducing color distribution. However, the design of a codebook is considered to be a challenging problem.
Discrete cosine transform/3DCT	Rasti et al. [36] and Qiao et al. [17]	One of the traditional compression techniques. It can remove intra-band correlation, but inter-band correlation is left untreated. Therefore, the 3D-DCT approach is introduced. It is an enhancement of DCT to extend the multiple-dimensional process level. However, the reconstructed image undergoes blocking effects, which severely degrade visual quality.
3D wavelets transforms	Tang and Pearlman [42], Shingate et al. [39], Zala and Parmar [41], and George and Manimekalai [40]	3D transform is not specifically designed for features of an HS image. It does not fully utilize the common background of inner bands and spectral redundancy.
Dictionary learning and prediction-based compression	Lin and Hwang [59], Castrodad et al. [56], Lin and Hwang [58], and Ülkü and Töreyin [55]	Dictionary-based learning methods take consideration of inherent characteristics of the input data. Uses only a few number of dictionary elements to present the dataset. It is a data-driving method. Using a dictionary-generating process is a key step to determining the compression performance.
Video coding H.264/HEVC	Gao et al. [62], Dusselaar et al. [63], and Podder et al. [76]	The HEVC standard has shown significant improvement over the state-of-the-art transformation-based still-image coding standard with spectral prediction modeling.

HS Library	Provider	Spectral Range	Description
AVIRIS—Airborne Visible Infrared Imaging Spectrometer	NASA	0.4–2.5 μm	AVIRIS is the most widely used HS database. AVIRIS data are basically related to the global environment and climate change; http://aviris.jpl.nasa.gov/.
HYDICE—Hyperspectral Digital Imagery Collection Experiment	Naval Research Laboratory	0.4–2.5 μm	HYDICE is a 206 channel imaging spectrometer. The data were collected on an ERIM CV 580 aircraft and a C-141 aircraft at altitudes up to 40,000 feet [78].
Natural scenes dataset	University of Manchester Foster, D. H., etc.	0.4–0.7 μm	The natural scenes dataset consists of a mixture of rural scenes from the Minho region of Portugal, containing rocks, trees, leaves, grass, and urban scenes from the cities of Porto and Braga; http://personalpages.manchester.ac.uk/staff/david.foster/Hyperspectral_images_of_natural_scenes_04.html.
Purdue’s Indianan Indian Pine test site	Purdue University Research Repository	0.4–2.5 μm	Purdue’s Indianan Indian Pine test site is another widely used HS image. It provides 12 band moderate dimension images and 220 band HS images; https://engineering.purdue.edu/~biehl/MultiSpec/hyperspectral.html.
Scene data	SCIEN, Stanford University	0.4–2.5 μm	The dataset contains images of faces, landscapes, and buildings. The images have been recorded with two HySpex line-scan imaging spectrometers, covering the spectral ranges of 0.4–1 μm and 1–2.5 μm; https://scien.stanford.edu/index.php/hyperspectral-image-data/.
USGS Digital Spectral Library	USGS Spectroscopy Laboratory	0.4–2.5 μm	USGS is a very useful laboratory data library, providing various remote sensing spectrometers; https://speclab.cr.usgs.gov/spectral-lib.html.

Abstract

Cited By

Figures (4)

Tables (3)

Equations (1)

Journal of the Optical Society of America A