IEEE Geoscience and Remote Sensing Magazine - March 2019 - 47

3

3
150

2
1

2
1

100

0

0

50

-1

-1
0

-2

-50

-3
(a)

-2

(b)

-3
(c)

FIGURE 8. (a) An unfiltered interferogram of Figure 2(a). (b) The PU result of the Kalman-filter-based PU method. (c) The rewrapped phase
fringes of (b).

} (s)

(8)

s

where g ($) is the generalized probability distribution function over } (s) and { (s).
There are many statistics-based methods [62]-[67].
The significant advantage of such methods is that the formulation shown in (8) can effectively fuse many kinds
of PU prior information, e.g., SAR image intensity or interferogram coherence. It is worth mentioning that the
SNAPHU method proposed in [63] is one of the most
representative statistics-based methods; it uses a statistical-cost, network-flow model to compute the most likely
PU solution. Some practical InSAR software (e.g., ISCE,
SNAP [68], and StaMPS [69]) use SNAPHU for their PU
processing step.
INTEGRATED DENOISING AND UNWRAPPING
METHODS
In the traditional InSAR processing flow, phase denoising and
unwrapping are two separate steps (i.e., denoising first and
then unwrapping). However, researchers noticed that such
separate steps inevitably introduce some extra processing errors because each step has different algorithm approximation
and presumption. Separately improving the algorithms of
denoising and unwrapping will result in the PU step unavoidably suffering from the algorithm errors of the denoising step.
To overcome this issue, in recent years, some integrated
denoising and unwrapping methods [24], [70]-[74], in which
PU performs simultaneously with the phase noise filtering,
have been of increasing interest. The integrated denoising
and unwrapping methods try to recover the information of
the noisy pixels. However, at present, the integrated denoising and unwrapping methods cannot completely take the
march 2019

ieee Geoscience and remote sensing magazine

REVIEW OF MULTIBASELINE PHASE UNWRAPPING
MB PU is an exciting and growing technique in InSAR.
Figure 9 illustrates the statistics for journal and conference
publications on MB PU (from Web of Science). As Figure 9,
shows, work on MB PU has taken off in recent years: the significant increase in the number of papers reveals the rapid
surge of interest in MB PU.
Unlike SB PU, MB PU is well posed but not ill posed.
Because MB PU can take advantage of baseline diversity
to significantly increase the ambiguity intervals of interferometric phases, it completely overcomes the limitation
of the Itoh condition. MB PU does not need a high spatial

The Total Number of Publications

arg max % g (} (s) | { (s)),

place of the traditional cascade processing framework. The
integrated denoising and unwrapping methods still prefer
the input interferogram when it is slightly filtered. Also,
their execution time is relatively high.
Figure 8(a) is the unfiltered interferogram of Figure 2(a),
Figure 8(b) is the PU result of Figure 8(a) obtained by the
Kalman-filter-based PU method [75]-[77], and Figure 8(c)
illustrates the rewrapped phase fringes of Figure 8(b). Comparing Figure 8(a) and (c), we can see the filtering functionality of the integrated denoising and unwrapping methods.

16
14
12
10
8
6
4
2
0

19
9
19 5
9
19 7
9
20 9
0
20 1
0
20 3
0
20 5
0
20 7
0
20 9
1
20 1
1
20 3
1
20 5
17

statistics-Based metHods
Statistics-based methods obtain the PU result by maximizing the conditional probability of the PU result under the
given wrapped phase. The generalized mathematical formulation of statistics-based methods is

FIGURE 9. Statistics for MB PU publications from 1995 to 2017.

47
IEEE Geoscience and Remote Sensing Magazine - March 2019

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