Professional Documents
Culture Documents
Singh 2018
Singh 2018
Abstract The prime objective is to harvest more and more information present
in a remotely sensed dark satellite image, captured under poorly illuminated cir-
cumstances. For imparting optimal quality enhancement, a recently proposed and
highly efficient Sine-Cosine optimizer is employed in association with a novel opti-
mally weighted gamma corrected (GC) fractional differential (FD) order masking
framework. Overall texture enhancement is achieved by optimally ordered FD mask-
ing along with its optimal augmentation with GC interim channel. Core objective
of entropy enhancement is fulfilled by keeping a proper check for over-enhanced or
saturated regions through the introduction of penalty term in the employed cost func-
tion, for adaptive exploration and identification of missing levels for more optimal
redistribution throughout the permissible range; so that natural look can be preserved
efficiently. Rigorous experimentation is performed by employing performance eval-
uation and comparison with preexisting highly appreciated quality enhancement
approaches.
1 Introduction
2 Proposed Methodology
HSI
Here, TRGB symbolizes the RGB to HSI equivalence conversion. Quality improve-
ment for color images is professionally achieved through isolated processing of lumi-
nance intensity channel only. Rest of the channels namely, hue and saturation are
kept preserved without any change. The weighted distributed summation enhanced
channel is finally achieved by amalgamation of statistically derived interim gamma
A Sine-Cosine Optimizer Based Gamma Corrected Adaptive … 635
Analytical formation for both interim images can be described as follows. Cumu-
lative distribution for input intensity channel is used to obtain gamma value set
optimally as
Making the core foundation of the “adaptive edge as well as texture augmentation”
principle, a highly robust self-proposed novel fractional differential mask framing
strategy is efficiently employed for constructive and collective “edge restoration
along with texture enhancement” which is also amalgamated and optimally associ-
ated with adaptively derived optimal gamma value set, through a parallel pipelining
for overall quality enhancement. Different types of fractional order differential 2-D
filters have been suggested till date, for textural quality improvements for an image.
Here, obeying the fractional calculus based standard Grunwald–Letnikov (G-L) def-
inition, vth order differential for f (x) (where f (x) ∈ [a, x], a ∈ R, x ∈ R), can
be symbolized as [15]
n−1
GL v −v −v
a Dx f (x) lim h −v f (x − mh) , where,
h→0
m0
m m
(−v) (−v + 1) . . . (−v + m − 1)
,
m! (5)
5 ∗ P C1 ; 3 ∗ Q C2 , (9)
[1, C1 , C2 ] 1, (−v), v 2 − v /2 , (10)
⎛ ⎞
(C2 /3) (C1 /5) (C1 /5)
⎜ ⎟
Hx ⎝ (C2 /3) (C1 /5) 1 ⎠; and
(C2 /3) (C1 /5) (C1 /5)
⎛ ⎞
(C2 /3) (C2 /3) (C2 /3)
⎜ ⎟
Hy ⎝ (C1 /5) (C1 /5) (C1 /5) ⎠, (11)
(C1 /5) 1 (C1 /5)
⎛
⎞
v 2 − v /6 (−v/5) (−v/5)
⎜
⎟
⎜ ⎟
Hx ⎜ v 2 − v /6 (−v/5) 1 ⎟; and
⎝
⎠
v 2 − v /6 (−v/5) (−v/5)
⎛
⎞
v 2 − v /6 v 2 − v /6 v 2 − v /6
⎜ ⎟
Hy ⎜
⎝ (−v/5) (−v/5) (−v/5) ⎟
⎠ (12)
(−v/5) 1 (−v/5)
For such masks, prescribed fractional order lies in the range (0, 1) and hence
optimal value according to image under consideration has to be evaluated adaptively.
Here, the linear filtering through mask Hx and Hy is done along left to right direction
across the rows followed by up to down across columns, respectively. Fractionally
ordered filtering is imparted through these masks to extract texture information as,
Iv Iin ⊗ H (13)
Adaptive behavior for various intensity levels can be implied by imparting hyper-
bolic profiled mapping derived by the adaptive gain adjustment parameter (λ) as
where
n ov ĩ mn < 0 ∪ ĩ mn > 1 , (17)
Here, X it stands for the current position of solution at tth iteration correspondingly
in the ith dimension. Here, randomness of the exploration phase is maintained by
employing four random variables, namely, r1 , r2 , r3 and r4 . The usage selection for
both of above equations is separately governed by r4 , and hence it is random and
equi-probable. In addition, Pit stands for ith dimension’s destination point for the
corresponding tth iteration. The above expressions behave somehow in a conjugate
638 H. Singh et al.
fashion and can be combined using the uniformly distributed random variable r4
in the range of [0, 1]. The above equation expresses the core updating mechanism
ensuring both exploration as well as exploitation collectively. The random parameter
r1 governs the direction of the movement. The direction can be inside or outside the
region in-between the solution and destination. Similarly, the parameter r2 stands for
the magnitude of the movement towards or outwards the destination. The random
parameter r3 imparts a random weightage for the destination so that the effect of des-
tination in defining the distance can be introduced stochastically. The corresponding
net effect is emphasized if r3 > 1, and deemphasized if r3 < 1. At last, the parameter
r4 is solely responsible for switching exchange in-between sine- and cosine-based
conjugate equations.
It can be easily understood that the entire interim space between two solutions
can be defined through the above equations. The relative as well as absolute change
of range of sine and cosine expressions leads to the relative updating of the posi-
tion outside/inside the interim region in-between itself and another solution. The
abovementioned random position characterized by r2 is defined in the range [0, 2π ]
employed through the updating equation. Hence, this mechanism ensures the col-
lective effect of exploration as well as exploitation for the entire search space for
the corresponding dimension iteratively. It is the promising intellectual behavior to
maintain balance of exploration and exploitation, which highlights the outperfor-
mance of the optimizer. Here, assuming a constant positive integral damping factor
(a), for the tth iteration, r1 is defined in a linearly reducing fashion as [16]:
where, t signifies the current iteration, and T stands for the total iteration count. The
resultant damping or range-reduction during the consecutive course of iterations
can be easily understood as an effect of r1 over the employed updating equation.
Also, it can be noticed the Sine-Cosine optimizer explores efficiently, when sine and
cosine function ranges in (1, 2] and [−2, 1). Correspondingly, the search space is
exploited efficiently when sine and cosine ranges in [−1, 1]. In the meantime, as the
iteration counter increases, the ranges of the sine and cosine functions are updated
to emphasize the exploitation of the available search space. The termination of this
optimizer is executed as the maximum iteration count is achieved. The entire process
flow for the proposed SCA-GCFDM approach can be summarized as shown in Fig. 1.
RGB
Here, THSI is HSI to RGB transformation process.
For explicit comparative numerical assessments, relevant indices are listed in Table 1.
640 H. Singh et al.
4 Conclusion
The proposed approach seems a more efficient method for information harvesting
through airborne remotely sensed dark satellite images, acquired under poor illumina-
tion. An efficient exploration of search space using Sine-Cosine optimizer leads to the
higher efficacy of the proposed textural enhancement along with proper edge restora-
tion. Thus, overall quality enhancement is imparted through optimal decision making
of the fractional differentiation order by employing this approach. Entire allowable
intensity span must be exploited so that the count of the void bins can be reduced and
hence, cost function is framed by introducing this void-bin count as a penalty term
along with desired assurance for entropy as well as contrast enhancement. Rigorous
experimentation is executed by employing the performance evaluation and compari-
son with preexisting recently proposed and highly appreciated quality enhancement
Table 1 Quantitative evaluation with comparison among input images [17–19], GHE [3], MMSICHE [5], ADAPHE [4], AVHEQ [6], AGCWD [9], HEOPC
642
[7], HEMIC [8], IEAUMF [13], and the proposed approach using various metrics termed as Brightness (B), Contrast (V), Entropy (H), Sharpness (S) and
Colorfulness (C)
S. No INDICES INPUT GHE MMISCHE ADAPHE AVHEQ AGCWD HEOPC HEMIC IEUMF Proposed
1 B 0.2573 0.5004 0.2955 0.4104 0.3193 0.4243 0.322 0.3849 0.3237 0.3994
V 0.0304 0.0859 0.0596 0.0623 0.0492 0.063 0.0461 0.052 0.0501 0.0798
H 6.8359 7.2603 7.0938 7.4255 7.0964 7.1282 6.9724 7.2701 7.0812 7.2519
S 0.299 0.513 0.3982 0.6406 0.3803 0.4353 0.3694 0.4034 0.462 0.6923
C 0.1264 0.267 0.1352 0.2081 0.156 0.2267 0.16 0.2016 0.1592 0.2314
2 B 0.106 0.5019 0.1573 0.3226 0.1253 0.3337 0.1294 0.2186 0.14 0.3899
V 0.0076 0.0846 0.0408 0.0431 0.012 0.0655 0.0102 0.015 0.0191 0.1195
H 5.5645 6.7583 6.0061 6.8852 5.749 6.547 5.7736 6.3054 5.8876 6.8342
S 0.2513 1.0145 0.4821 0.6922 0.318 0.8437 0.2956 0.406 0.4206 1.0147
C 0.0489 0.2549 0.0652 0.1576 0.0572 0.1586 0.0602 0.1113 0.0619 0.2275
3 B 0.2232 0.526 0.2581 0.3636 0.2893 0.3759 0.2651 0.3477 0.3006 0.4115
V 0.044 0.0669 0.0723 0.0881 0.0421 0.0855 0.0355 0.0485 0.0621 0.1263
H 5.9635 6.3469 6.3378 6.5588 7.2381 6.1279 7.1482 7.3685 7.2859 7.4428
S 0.6581 0.8567 0.8033 1.0538 0.5041 0.9919 0.5082 0.5637 0.7700 1.2690
C 0.1328 0.3077 0.1481 0.2312 0.1615 0.2414 0.1497 0.1984 0.1827 0.3231
(continued)
H. Singh et al.
Table 1 (continued)
S. No INDICES INPUT GHE MMISCHE ADAPHE AVHEQ AGCWD HEOPC HEMIC IEUMF Proposed
4 B 0.3519 0.501 0.3787 0.4778 0.6367 0.5218 0.4485 0.528 0.4671 0.7297
V 0.0094 0.0859 0.0339 0.0531 0.0544 0.0345 0.0183 0.0294 0.0573 0.0603
H 6.88 7.2077 7.0756 7.4709 7.6275 7.2984 7.2257 7.4517 7.5047 7.7535
S 0.2415 0.7182 0.4474 0.6778 0.5638 0.4524 0.337 0.4254 0.8688 1.0363
C 0.2126 0.3204 0.2288 0.3076 0.3684 0.3169 0.2687 0.3135 0.295 0.6587
5 B 0.0612 0.6079 0.1259 0.1842 0.2692 0.182 0.1235 0.4127 0.1234 0.2399
V 0.0071 0.0322 0.0542 0.0463 0.1188 0.0415 0.0281 0.0433 0.0305 0.0607
H 3.0107 3.8806 3.2498 4.0203 3.354 3.2902 3.8289 4.4715 3.34 4.6541
S 0.2812 0.6202 0.698 0.7238 1.1706 0.713 0.5694 0.6471 0.5945 0.8954
C 0.0875 0.5103 0.2138 0.2477 0.3768 0.2425 0.1658 0.455 0.1703 0.3743
6 B 0.1446 0.526 0.1752 0.3212 0.1939 0.3186 0.1811 0.4248 0.1871 0.4314
V 0.032 0.0668 0.0569 0.0735 0.057 0.0765 0.0489 0.0501 0.0569 0.1787
H 5.2964 6.1069 5.6815 6.2404 5.4922 5.7146 5.586 6.3627 5.488 6.3487
S 0.4526 0.8683 0.561 0.9733 0.6203 0.9196 0.5558 0.6105 0.666 1.2664
C 0.1228 0.3483 0.1602 0.2345 0.1672 0.2513 0.1521 0.2781 0.1618 0.3732
A Sine-Cosine Optimizer Based Gamma Corrected Adaptive …
643
644 H. Singh et al.
References
1. Singh, H., Kumar, A., Balyan, L.K., Singh, G.K.: A novel optimally weighted framework of
piecewise gamma corrected fractional order masking for satellite image enhancement. Comput.
Electr. Eng. 1–17 (2017). https://doi.org/10.1016/j.compeleceng.2017.11.014
2. Singh, H., Kumar, A., Balyan, L.K., Singh, G.K.: Swarm intelligence optimized piecewise
gamma corrected histogram equalization for dark image enhancement. Comput. Electr. Eng.
1–14 (2017). https://doi.org/10.1016/j.compeleceng.2017.06.029
3. Gonzalez, R.C., Woods, R.E.: Digital image processing, 3rd edn. (2007)
4. Zuiderveld, K.: VIII.5. Contrast limited adaptive histogram equalization. Graph. Gems 474–485
(1994).https://doi.org/10.1016/b978-0-12-336156-1.50061-6
5. Singh, K., Kapoor, R.: Image enhancement via median-mean based sub-image-clipped his-
togram equalization. Optik (Stuttg) 125, 4646–4651 (2014)
6. Lin, S.C.F., Wong, C.Y., Rahman, M.A., et al.: Image enhancement using the averaging his-
togram equalization approach for contrast improvement and brightness preservation. Comput.
Electr. Eng. 46, 356–370 (2014)
7. Wong, C.Y., Jiang, G., Rahman, M.A., Liu, S., Lin, S.C.F., Kwok, N., Shi, H., Yu, Y.H., Wu,
T.: Histogram equalization and optimal profile compression based approach for colour image
enhancement. J. Vis. Commun. Image Represent. 38, 802–813 (2016)
8. Wong, C.Y., Liu, S., Liu, S.C., Rahman, A., Lin, C., Jiang, G., Kwok, N., Shi, H.: Image contrast
enhancement using histogram equalization with maximum intensity coverage. J. Modern Opt.
63(16), 1618–1629 (2016)
9. Huang, S.C., Cheng, F.C., Chiu, Y.S.: Efficient contrast enhancement using adaptive gamma
correction with weighting distribution. IEEE Trans. Image Process. 22, 1032–1041 (2013)
10. Singh, H., Kumar, A.: Satellite image enhancement using beta wavelet based gamma corrected
adaptive knee transformation. In: IEEE International Conference Communication and Signal
Processing (ICCSP), India, pp. 128–132. https://doi.org/10.1016/j.compeleceng.2017.06.029
(2016)
11. Singh, H., Agrawal, N., Kumar, A., Singh, G.K., Lee, H.N.: A novel gamma correction approach
using optimally clipped sub-equalization for dark image enhancement. In: IEEE International
Conference Digital Signal Process (DSP) 2016, China, pp. 497–501. https://doi.org/10.1109/i
cdsp.2016.7868607 (2016)
12. Singh, H., Kumar, A., Balyan, L.K., Singh, G.K.: A novel optimally gamma corrected intensity
span maximization approach for dark image enhancement. In: IEEE International Conference
Digital Signal Process (DSP) 2017, London, United Kingdom, pp. 1–5. https://doi.org/10.110
9/icdsp.2017.8096035 (2017)
13. Lin, S.C.F., Wong, C.Y., Jiang, G., Rahman, M.A., Ren, T.R., Kwok, N., Shi, H., Yu, Y.H.,
Wu, T.: Intensity and edge based adaptive unsharp masking filter for color image enhancement.
Optik (Stuttg) 127, 407–414 (2016)
14. Singh, H., Kumar, A., Balyan, L.K., Singh, G.K.: Regionally equalized and contextually clipped
gamma correction approach for dark image enhancement. In: IEEE International Conference
Signal Process. Integrated Networks (SPIN), India, pp. 431–436. https://doi.org/10.1109/spin.
2017.8049988 (2017)
15. Pu, Y.F., Zhou, J.L., Yuan, X.: Fractional differential mask: a fractional differential-based
approach for multiscale texture enhancement. IEEE Trans. Image Process. 19(2), 491–511
(2010)
16. Mirjalili, S.: SCA: a sine cosine algorithm for solving optimization problems. Knowl.-Based
Syst. 96, 120–133 (2016)
A Sine-Cosine Optimizer Based Gamma Corrected Adaptive … 645
17. NASA Visible Earth: Home. http://visibleearth.nasa.gov/ (2017). Accessed on 2 June 2017
18. SATPALDA. http://www.satpalda.com/gallery/ (2017). Accessed on 2 June 2017
19. CRISP. http://www.crisp.nus.edu.sg/ (2017). Accessed on 2 June 2017