IEEE Geoscience and Remote Sensing Magazine - March 2023 - 56

0.2
0.15
0.1
0.05
5
4
3
2
1
(a)
(b)
(c)
(d)
(e)
(f)
FIGURE 21. The fusion results of five different HS-MS fusion methods: (a) the reference image, (b) Blind-STEREO, (c) CSTF, (d) LTMR,
(e) LTTR, and (f) SC-LL1. MRAE: mean relative absolute error.
happens in practice. Taking the interimage variability
phenomenon into consideration when modeling the
degradation process is a key challenge for future research.
HYPERSPECTRAL UNMIXING
Owing to its acquired continuous abundance maps, SU has
widely solved the inversion problems of typical ground object
parameters, such as the vegetation index, surface temperature,
and water turbidity, in the past several decades
[196], [197], [198] and has been successfully applied in
some RS applications, such as forest monitoring and land
cover change detection [199]. In addition, due to the mixing
phenomenon caused by heterogeneity and the stratified
distribution of ground objects, SU can effectively realize
crop identification and monitoring [200], [201], [202].
TABLE 11. A QUANTITATIVE COMPARISON OF DIFFERENT
METHODS FOR HS-MS FUSION.
INDEX
BLIND-STEREO CSTF
MPSNR 53.49
ERGAS
SAM
RMSE
CC
0.3584
1.0132
0.0027
0.9993
54.28
0.3317
0.8841
0.0025
0.9994
LTTR
38.11
1.9761
3.7971
0.0165
0.9819
LTMR
39.58
1.5851
3.0495
0.0132
0.987
SC-LL1
54.13
0.3076
0.8213
0.0023
0.9995
When the mixing scale is macroscopic and each incident
light-reaching sensor has interacted with just one material,
the measured spectrum is usually regarded as a linear
mixing, as described in Figure 22(a). However, due to the
existence of nonlinear interactions in real scenarios, several
physics-based approximations of the nonlinear linear
mixing model (NLMM) have been proposed, mainly covering
two types of mixing assumptions: intimate mixture
[Figure 22(b)] and multilayered mixture [Figure 22(c)]. The
former describes the interactions suffered by a surface composed
of particles at a microscopic scale. Intimate mixture
usually occurs in scenes containing sand and mineral mixtures
and requires a certain kind of prior knowledge of the
geometric positioning of the sensor to establish the mixture
model. The latter characterizes the light reflectance of
various surface materials at a macroscopic scale. Multilayered
mixture usually occurs in scenes composed of materials
with height differences, such as forests, grassland, and
rocks, containing many nonlinear interactions between
the ground and canopy. In general, a multilayered mixture
consisting of more than two orders is ignored owing to its
negligible interactions. For the second-order multilayered
mixture model, the family of bilinear mixing models is
usually adopted to solve the NLMM. Due to the low spatial
resolution of sensors, many pixels mixed with different
pure materials exist in HS imagery, which inevitably
56
IEEE GEOSCIENCE AND REMOTE SENSING MAGAZINE MARCH 2023
SAM
MRAE
Zoomed
Output
IEEE Geoscience and Remote Sensing Magazine - March 2023

Table of Contents for the Digital Edition of IEEE Geoscience and Remote Sensing Magazine - March 2023
