IEEE Geoscience and Remote Sensing Magazine - June 2019 - 147

the information content indefinitely. Figure 9 shows the
R 2 correlation matrix computed on the I D image of the
Benton County Hyperion data set. R 2 represents the correlation of each SCV band with the other ones. Four main
correlated band groups can be observed, and the adjacent
bands are highly similar. This indicates the redundancy
present in the considered HS data set. When we look for
changes in the multitemporal domain, we expect more redundancy in the HS change vectors than in the spectral
vectors of each single image.
Noise generated by both external sources (atmospheric
effects due to absorption and scattering) and internal sources (instrument noise) also affects HS sensors. These noise
sources are also present in MS sensors. The atmospheric haze
can be mitigated with atmospheric correction techniques,
such as atmospheric removal (i.e., ATREM) [42] or fast lineof-sight atmospheric analysis of spectral hypercubes (i.e.,
FLAASH) [43] algorithms. Instrument noise is introduced
by 1) nonideal electronics (thermal and shot noise); 2) quantization; 3) malfunctioning sensors, which generate impulsive noise or missing pixels/lines; and 4) striping noise due
to a calibration error of the array of sensors in push-broom
scanners [11]. When the spectral resolution increases, the
measured reflectance at the sensor decreases, whereas the
signal-independent noise, such as the thermal noise, does
not, thus decreasing the signal-to-noise ratio [44]. When the
number of bands increases above a certain limit, information represented by the HS bands becomes more sparse and
implicit, which may degrade the discriminative ability of a
CD, especially in identifying multiclass changes.

section. In this section, we provide an overview of the
main techniques proposed in the literature for the different categories.
ANOMALOUS CHANGE DETECTION
ACD aims at identifying anomalous changes between
images by suppressing background and accentuating
changes [76]. In general, ACD approaches for HS images
can be separated into single-instance and multi-instance
target detection [56]. The key point is to investigate the
statistics of the images, increase the detection probability of changes caused by human activity, and suppress
background in the image scene sequences. For ACD, special attention is usually paid to the detection of small
changes resulting from the insertion, deletion, or movement of manufactured small objects as well as from small

TABLE 2. SUMMARY OF MAIN TECHNIQUES FOR CD IN HS
DATA (THE TABLE IS NOT EXHAUSTIVE).
APPLICATION

TECHNIQUE

REFERENCE

ACD

Chronochrome

Schaum and Stocker [45]

Covariance equalization

Schaum and Stocker [46]

Clustering

Shimoni et al. [47]

Statistical modeling

Theiler et al. [48]
Acito et al. [49]

Reed-Xiaoli algorithm

Wu et al. [50]
Zhou et al. [51]

Model based

Meola et al. [52]

Error and noise mitigation

Theiler and Wohlberg [53]
Vongsy et al. [54]

CHANGE DETECTION TECHNIQUE IN
MULTITEMPORAL HYPERSPECTRAL IMAGES
Table 2 shows a nonexhaustive summary of existing CD
methods in HS images according to the categorization related to the CD application presented in the "Categories of
Change-Detection Approaches for Hyperspectral Images"

Meola et al. [55]
Meola and Eismann [56]
Eismann et al. [57]
Binary CD

Transformation

Frank and Canty [58]
Nielsen [59], [60]
Ortiz-Rivera et al. [61]
Hemissi et al. [62]
Liu et al. [63]
Wu et al. [64]

1

Unmixing

Du et al. [65]
Ertürk and Plaza [39]
Ertürk and Plaza [66]
Ertürk et al. [67]

Others

Yuan et al. [68]
Du et al. [69]
Chen and Wang [70]

Multiclass CD

Time-series CD

0

Clustering

Liu et al. [35]

Compressed CVA

Liu et al. [36], [37]

Band selection

Liu et al. [28]

Unmixing

Liu et al. [38]

Binary SCVs

Marinelli et al. [71]

Clustering

Yang and Sun [72]

Unmixing

Hemissi et al. [73]
Henrot et al. [74]

FIGURE 9. The R2 correlation matrix of the ID image, where four

Veganzones et al. [75]

highly correlated adjacent band groups indicate the redundancy.

JUNE 2019

IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE

147
IEEE Geoscience and Remote Sensing Magazine - June 2019

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - June 2019
