Comprehensive framework for optimization of

overlapping pixels in optically butted focal planes

Ajay Kumar Krishnakumar ,* Nitesh Thapa, Narasiha B. Sharma ,

and Arti Sarkar
Indian Space Research Organization, Space Applications Centre, Ahmedabad,
Gujarat, India

Abstract. Very high-resolution space borne missions demand finer spatial sampling and larger
swath coverage for several commercial and strategic applications including cartography, disaster
monitoring, urban planning, and surveillance. To meet these contrasting requirements, optical
sensors often employ optical butting techniques in their focal planes as this enables the usage of
small format detectors instead of large format single detectors, thereby reducing fabrication
costs. In the optical butting technique, small reflecting mirrors placed before the focal plane
split the optical field into smaller segments, which are alternately imaged by individual small
format detectors. A single continuous image is formed using the small image segments through
image processing technique. This requires optimal overlaps among the image segments with
adequate image quality in the common imaged region. Factors, such as the geometric alignment
of detectors with the butting mirror edge, edge-vignetting effects causing modulation transfer
function, and signal-to-noise ratio degradations in the overlapping regions, largely determine the
achievable overlap and image quality in the optically butted focal planes. Our study presents
a comprehensive framework for the optimization of overlapping pixels in the optically butted
focal planes based on a quantitative multicriteria analytical approach. The proposed framework
has been tested for various sensor configurations to establish its efficacy. Our study will enable
the efficient design and development of high-performance focal plane assemblies for high-
resolution optical sensors. © 2023 Society of Photo-Optical Instrumentation Engineers (SPIE)
[DOI: 10.1117/1.OE.62.2.023103]
Keywords: optical butting; field splitting; relative illumination; signal-to-noise ratio; modula-
tion transfer function.
Paper 20220785G received Aug. 31, 2022; accepted for publication Feb. 6, 2023; published
online Feb. 28, 2023.

1 Introduction
Very high-resolution spaceborne optical imaging sensors have found potential applications in
several commercial and strategic applications including cartography, urban development, disas-
ter monitoring and management, and surveillance. Over the years, optical sensors have improved
both spatial resolution and swath coverage1–3 due to advancements in charge coupled device
(CCD) and complementary metal oxide semiconductor (CMOS) image sensor technologies.
However, due to detector fabrication technology limitations, accommodating many pixel
elements in a single chip is very challenging and involves a high development cost.3
Field splitting techniques with multiple small format detectors placed in the focal plane, such
as optical butting,1,4,5 have emerged as an alternative approach to overcome the limitations for
covering large swath. In this technique, multiple reflective mirrors are placed in the path of the
focusing light beam to split the field into small segments for imaging by individual small format
detectors placed in the orthogonal or different planes. The individual image strips are then
mosaicked to generate a continuous across track image. This technique has been widely used
in IKONOS-2, Pleiades, KOMPSAT-2, and some payloads flown by ISROs.2,5–9

*Address all correspondence to K. Ajay Kumar, [email protected]

0091-3286/2023/$28.00 © 2023 SPIE

Optical Engineering 023103-1 February 2023 • Vol. 62(2)

Ajay Kumar et al.: Comprehensive framework for optimization of overlapping pixels in optically butted focal. . .

The seamless mosaic can only be generated if an optimum number of pixels imaging the
same location (generally referred as overlapping pixels) with good image quality in terms of
modulation transfer function (MTF) and signal-to-noise ratio (SNR).10 However, the design and
realization of a focal plane assembly (FPA) having multiple detectors butted together is a very
complicated process and several factors can significantly affect the achievable pixel overlap.11
Fuqiang et al. reported some of these as design errors, fabrication and alignment errors in butting
mirrors, alignment errors in the main telescope, detector placement errors, and thermal distor-
tions. If not addressed properly, these multiple errors can result in non-optimal overlapping pix-
els, causing either potential image gaps (underlaps) or reduction in swath (due to excess overlap).
Additionally, the image quality of the pixels in the overlapping region may also get severely
degraded, affecting the final mosaicked image quality. The image quality degradations generally
emanate from the butting mirror edges coming in the path of the optical beam that causes
vignetting of light rays resulting in poor MTF and SNR performance in the vignetting region.
Given this, it is essential to ensure optimum overlap between image segments.
Fuqiang et al.11 presented a method for calculating overlapping pixels with MTF assurance in
the overlap region and not considered SNR performance degradations for optimal overlap pixels.
This study proposes a generic and comprehensive framework for optimizing overlap pixels based
on a multicriteria quantitative analytical approach that includes both MTF and SNR perfor-
mances. A mathematical model for SNR degradation criteria is developed and effectively used
in the framework. The framework has been extensively test validated through three case studies
demonstrating its efficacy. This study presents a comprehensive methodology for the design and
development of optically butted focal plane assemblies for future high-resolution space borne
missions.
The remainder of this paper is organized as follows. Section 2 elaborates on the concept of
optical butting in focal plane assemblies, introduces various taxonomies subsequently used in the
paper, and discusses various factors that influence the overlapping pixels. In Sec. 3, the proposed
framework is presented including the theoretical basis and implementation aspects. A model for
the SNR criteria is developed and discussed in this section. In Sec. 4, the efficacy of the frame-
work is demonstrated by presenting the results of three case studies. Conclusions are drawn
in Sec. 5.

2 Optical Butting Technique and Overlap Requirements

Optical butting is a method for virtually generating a larger format detector using multiple small
format detectors by folding light from different across track fields into two planes. Theoretically,
there is no limit on the number of detectors that can be butted using this technique. An optical
butting scheme for four short format detectors is shown in Fig. 1. The rays coming from the
telescope will focus on the direct detectors and those reflected from the butting mirror will focus
on the detectors in the orthogonal plane as shown in Fig. 1. The butting mirror edge causes
vignetting of fields as it intersects only a portion of the optical beam corresponding to that field.

Fig. 1 Optical schematic showing optical butting.

Optical Engineering 023103-2 February 2023 • Vol. 62(2)

Ajay Kumar et al.: Comprehensive framework for optimization of overlapping pixels in optically butted focal. . .

Fig. 2 Isometric view of optical butting scheme for a telecentric system.

At the mirror edge, a portion of the beam corresponding to a particular field is reflected
and the remaining portion passes directly. This leads to the formation of two images of the same
field; one in the reflected plane (i.e., reflected by butting mirror) and the other in the direct plane.
These two images will have gradually reducing intensity due to partial beam participation as
depicted in Fig. 1. The region where the beam is partially reflected and partially transmitted
will be referred to as the vignetting region. The extent of the vignetting region depends on the
f∕# of the optical system and the distance d between the butting mirror and focal plane as shown
in Fig. 2. For telecentric systems, the extent of vignetting region L is given in the following
equation:

d
L¼
EQ-TARGET;temp:intralink-;e001;116;400 : (1)
f∕#

Signal in the reflected plane of butting region is proportional to the area of beam that gets
incident on the butting mirror. Similarly, signal in direct plane is proportional to the area of beam
that is not incident on the butting mirror. The area of the beam not intersected by butting mirror
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
can be worked by obtaining area under the ellipse (i.e., 2 ∫ x−L b 1 − 4 Lt 2 dt) having semi-minor
2

2
axis L∕2 and semi-major axis b. Using the area under the ellipse formula, the proportion of
signal received in direct (Sd ðxÞÞ and reflected plane ðSr ðxÞÞ for a given field position x
(−L
2 ≤ x ≤ 2 ) with respect to total incident signal (ST ) can be derived as per the following
L

equation:
Rx qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
b 1 − 4 Lt 2 dt
2
Sd ðxÞ 2 −L2
EQ-TARGET;temp:intralink-;e02a;116;249 ¼ R L
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; (2a)
ST 2 2
b 1 − 4 Lt 2 dt
2

−L2

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi3 2 Rx
1 − 4 Lt 2 dt7
2
Sr ðxÞ 6 −L2
b 2
EQ-TARGET;temp:intralink-;e02b;116;175 ¼ R41 − R L qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi5: (2b)
ST 2 −2 L b 1 − 4 Lt 2 dt
2

By solving Eqs. 2(a) and 2(b), following equations provide estimation of Sd ðxÞ and Sr ðxÞ:
 rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  
Sd ðxÞ 1 8 x L2 L2 2x
¼ f d ðxÞ ¼ þ − x2 þ sin−1 ; (3)
2 πL2
EQ-TARGET;temp:intralink-;e003;116;120

ST 2 4 8 L

Optical Engineering 023103-3 February 2023 • Vol. 62(2)

Ajay Kumar et al.: Comprehensive framework for optimization of overlapping pixels in optically butted focal. . .

Fig. 3 Relative illumination profile in both the direct and reflected planes in the vignetting region.

  rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  
Sr ðxÞ 1 8 x L2 L2 −1 2x
¼ f r ðxÞ ¼ R − 2 − x þ sin
2 ; (4)
2 πL
EQ-TARGET;temp:intralink-;e004;116;412

ST 2 4 8 L

where R is the reflectance of the butting mirror and the functions, f d and f r are one to one
functions. It can be seen from these equation that SdSðxÞ
T
and SrSðxÞ
T
have become independent of
semi-major axis.
Using Eqs. (3) and (4), relative illumination profiles of direct and reflected planes in the
vignetting region are shown in Fig. 3. In principle, it is always preferred to minimize the
vignetting region to reduce the portion of the image, in which correction is to be performed.
Possible detector placement scenarios are shown in Figs. 4(a)–4(c). In the first scenario, as
shown in Fig. 4(a), detectors are placed such that they form a contiguous image with no overlap.
In Fig. 4(b), detectors are placed with a discontinuity (underlap) this will lead to gaps in the
image, whereas in the third scenario [Fig. 4(c)], some pixels in both detectors see a common
region (overlap). These pixels share the same field from the telescope, however, are exposed to
different illumination profiles resulting in poor SNR and MTF performance. However, such
overlaps are essential for seamless mosaicking of the images as it helps in restoring SNR per-
formance of the pixels in the vignetting region.9 Hence, it is essential to optimize overlap pixels
in the vignetting region accounting for both SNR and MTF performances. The comprehensive
framework proposed in this study helps in determining optimal overlapping pixels to ensure
a seamless image mosaicking process.

3 Comprehensive Framework for Optimal Selection of Overlapping

Pixels
This section deals with the comprehensive framework and its implementation aspect in detail.
The proposed framework is based on a multicriteria quantitative analytical approach that helps
determine optimal overlapping pixels based on the alignment uncertainty of optical components,
MTF, and SNR degradation criteria.

Optical Engineering 023103-4 February 2023 • Vol. 62(2)

Ajay Kumar et al.: Comprehensive framework for optimization of overlapping pixels in optically butted focal. . .

Fig. 4 Detectors are placed to form (a) a contiguous image, (b) underlap, and (c) such to form
an overlap region.

3.1 Alignment Uncertainty of Optical Components

To have a contiguous image (without any overlap), edge of the butting mirror and first pixel of
both detectors must be aligned. This constraint complicates the alignment procedure as exact
locations may not be known accurately. The placement of FPA elements with respect to the exit
pupil will affect the length of the vignetting region. To ensure overlap between two detectors,
the total uncertainty of the image plane concerning the telescope field needs to be accounted for.
The errors due to alignment uncertainty of optical components can be referred to as AUOC and
is given by the following equation:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
AUOC ¼ 2  Δ2b þ Δ2d þ Δ2o ;
EQ-TARGET;temp:intralink-;e005;116;291 (5)

where Δb are the alignment errors associated with estimating position of the edge of the butting
mirror. This shall also include all the errors occurred during fabrications and realization.
Δd are the alignment errors associated with placement of the detector. Δo are other errors, such
as distortions and estimation of vignetting region.
Several innovative techniques for alignment and integration process of FPA have been
reported in the literature to achieve submicron level alignment accuracies, such as adjustable
mechanisms are used for mounting of components in FPA modules,4 usage of special ground
equipment viz. classical coordinate measuring machines,1 and active alignment techniques for
optically butted FPAs.5 These techniques significantly help in minimizing these errors.

3.2 Signal-to-Noise Ratio Criterion

In the butting region, energy spreads in two different pixels, hence, the addition of the signal
afterward does not recover the SNR completely due to increased read noise associated with the
two pixels. Based on SNR degradation, a mathematical model is worked out to define the criteria
for selecting the number of overlap pixels.

Optical Engineering 023103-5 February 2023 • Vol. 62(2)

Ajay Kumar et al.: Comprehensive framework for optimization of overlapping pixels in optically butted focal. . .

Total noise in optical imaging systems comprises of mainly photon noise and dark noise
(i.e., noise when there is no input signal). Total signal after combining Sd ðxÞ and Sr ðxÞ is given
pffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffi
by Sc. The photon noise associated with it is for the total signal. Sd ðxÞ, Sr ðxÞ, and Sc are
the photon noise associated with signals in direct, reflected planes, and combined signal, respec-
tively. Here it is considered that the detector Dd receives direct signal, whereas the detector Dr
receives reflected signal. Dark noise associated with the detector Dd and Dr is N d and N r ,
respectively. The total dark noise of the combined signal in the vignetting region is given in
Eq. (6) assuming that dark noises associated to these detectors are random:
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Nc ¼
EQ-TARGET;temp:intralink-;e006;116;637 N 2d þ N 2r : (6)

SNRs in the vignetting region in Dd and Dr detectors are given by the following equation:

Sd
SNRd ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi; (7)
Sd þ N 2d
EQ-TARGET;temp:intralink-;e007;116;585

Sr
SNRr ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi : (8)
Sr þ N 2r
EQ-TARGET;temp:intralink-;e008;116;527

SNR after the addition of signals from two pixels imaging the same field is expressed in the
following equation:

Sd þ Sr SC
SNRc ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi : (9)
Sd þ Sr þ N d þ N r SC þ N 2c
EQ-TARGET;temp:intralink-;e009;116;477

2 2

SNR criterion is defined as the condition where the SNR after addition of signals from two
different planes, i.e.,SNRc , is always greater than the SNR of individual planes, i.e., either
SNRd or SNRr. So the condition for the direct plane can be mathematically represented by the
following equation:

SC Sd ðxÞ SC Sd ðxÞ
SNRc ≥ SNRd ⇒ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ≥ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi⇒ ≥ : (10)
SC þ N c Sd ðxÞ þ N d SC þ N c 2 Sd ðxÞ þ N d 2
EQ-TARGET;temp:intralink-;e010;116;383

2 2

This equation can be rearranged to form a quadratic equation as given in the following
equation:
  2  
Nc2 Sd ðxÞ Sd ðxÞ N 2
EQ-TARGET;temp:intralink-;e011;116;311 1þ − − d ≤ 0: (11)
SC SC SC SC

Solving Eq. (11) leads to Eq. (12):

rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 ffi
N2 2
1þ 1 þ 4 SCd 1 þ NSCc
Sd ðxÞ
EQ-TARGET;temp:intralink-;e012;116;253 ≤  2
 : (12)
SC 2 1 þ NSCc

Considering ∼0.97 reflectivity of the butting mirror (i.e., R ≈ 1), SC will be nearly equal to
ST as ST ¼ Sd ðxÞ þ SrRðxÞ and SC ¼ Sd ðxÞ þ Sr ðxÞ. Therefore, SdSðxÞC
can be equated with relative
illumination as given in Eq. (3). The fields that satisfy Eq. (12) in direct plane will have improve-
ment in SNR if the signal from the reflected plane is added to it. Assuming, xr and xd as the
distance of the field from butting mirror edge in the focal plane from where detector Dr and Dd
will start imaging. Hence, by numerically solving Eqs. (3) and (12), xr and xd can be estimated as
represented in Eqs. (13a) and (13b):

Optical Engineering 023103-6 February 2023 • Vol. 62(2)

Ajay Kumar et al.: Comprehensive framework for optimization of overlapping pixels in optically butted focal. . .

0 ffi1
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

N 2d N 2c
B1 þ 1 þ 4 SC 1 þ SC C
B   C
xr ¼ f −1
d B C; (13a)
@ N 2c A
EQ-TARGET;temp:intralink-;e13a;116;735

2 1 þ SC

ffi1
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 0
N 2r N 2c
B 1 þ 1 þ 4 SC 1 þ SC C
−1 B   C
xd ¼ f r B C; (13b)
@ N 2c A
EQ-TARGET;temp:intralink-;e13b;116;655

2 1 þ SC

where f −1 −1
d and f r are inverse functions as given in Eqs. (2) and (3). Overlap region defined by
this criterion can be expressed as shown in Eq. (14), where OSNR is overlap arrived at using
SNR criteria:

OSNR ðSc ;N d ;N r Þ ¼ jxr j þ jxd j

EQ-TARGET;temp:intralink-;e014;116;562

0 ffi1
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
 0 1
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

N2 2 N 2r N 2c
B1 þ 1 þ 4 Scd 1 þ NScc C B 1 þ 1 þ 4 Sc 1 þ Sc C
B   C B   C
¼ f −1
d B Cjþjf −1
r B C : (14)
@ 2
2 1 þ NScc A @ 2 1 þ Sc N 2c A

Condition I. When the read noise of both the detectors is insignificant compared to the
N 2d N 2r N 2c
photon noise, i.e., when the system is photon noise limited, i.e., Sc ; Sc , and Sc → 0 then
limN2 →0 SdSðxÞ
c
; SrSðxÞ
c
¼ 1 therefore OSNR ðSc ; N d ; N r Þ ¼ jf −1 −1
d ð1Þj þ jf r ð1Þj ¼ L. This means that
Sc

the detector shall be placed such that the entire signal (reflected or transmitted) shall be collected
as shown in Fig. 5.
Condition II. When read-out noise is higher than photon noise, i.e., for electronics
N 2d N 2r
noise limited performance system, Sc ; Sc , and N 2c
Sc → ∞ then limN2 →∞ SdSðxÞ
c
¼ NNdc and
Sc

limN2 →∞ SrSðxÞ
c
¼ NNcr . If both detector performances are identical, then limN2 →∞ SdSðxÞ
c
; SrSðxÞ
c
¼ 0.7.
Sc Sc

Therefore, OSNR ðST ; N d ; N r Þ ¼ jf −1 −1

d ð0.7Þj þ jf r ð0.7Þj. Solving this numerically results in
−1 −1
jf d ð0.7Þj ¼ 6.25 and jf r ð0.7Þj ¼ 7 . Hence, OSNR ðSc ; N d ; N r Þ ¼ 0.3L as shown in Fig. 6.
L L

Fig. 5 SNR in the vignetting region for a photon noise limited system.

Optical Engineering 023103-7 February 2023 • Vol. 62(2)

Ajay Kumar et al.: Comprehensive framework for optimization of overlapping pixels in optically butted focal. . .

Fig. 6 SNR in the vignetting region for electronics noise limited system.

Considering the above discussions, OSNR ðSc ; N d ; N r Þ is bounded function between given by
the following equation:

EQ-TARGET;temp:intralink-;e015;116;477 0.3L ≤ OSNR ðSc ; N d ; N r Þ ≤ L: (15)

3.3 Modulation Transfer Function Criterion

MTF of an optical system is defined as modulus of autocorrelation of its aperture function12
as given in Eq. (16), where Ax ðu; vÞ is the aperture function for a given field f ov ðxÞ:
ZZ
MTFðf u ; f v ; f ov ðxÞÞ ¼ gfu fv ðxÞ ¼
EQ-TARGET;temp:intralink-;e016;116;384 Ax ðu; vÞAx ðu − f u ; v − f v Þdudv ; (16)

where f u ; f v are the spatial sampling frequency in the orthogonal direction. The exit pupil of the
telescope system usually defines the aperture function, and the apodization function describes
the intensity and phase changes due to optical aberration. In optical butting, the aperture function
gets modified with field due to vignetting of the optical beam by the butting mirror. Cutoff
frequency (f cutoff ) of an optical system is given by 1∕ðλf#Þ, and the Nyquist frequency (f Ny )
of the optical system is half of reciprocal to the detector pixel size.
To understand the drop in MTF in the vignetting region, a diffractive limited system with
circular aperture is considered. Due to the butting mirror, the exit pupil gets modified into seg-
ments of different areas across its field for reflected and direct imaging planes. Vignetting will
lead to a drop in MTF and can be computed numerically for different spatial frequencies using
Eq. (16). The drop in MTF (in direction of vignetting) as a function of vignetting by butting
mirror (i.e., relative illumination) is given in Fig. 9. The MTF drop is computed for four sampling
conditions (different ratios of sampling frequency to optics cutoff frequency). As the vignetting
increases, MTF degrades for a given field. MTF drop in the vignetting region varies with f Ny of
the imaging system. In Fig. 7(a), relative MTF for both the planes at center field is around 70%,
i.e., for f Ny of 0.2f cutoff . As the f Ny increases, the drop in MTF also increases in both the plane
for a given field position. In Figs. 7(c) and 7(d), relative MTF drops to zero much before the
center field for imaging systems with f Ny of 0.6f cutoff and 0.8f cutoff . Generally, high-resolution
systems are realized with limited aperture and smaller pixels to meet the size and mass require-
ments. For such high-resolution systems, the MTF at f Ny can be as low as zero in the vignetting
region. However, it can be recovered using image sharpening techniques provided adequate

Optical Engineering 023103-8 February 2023 • Vol. 62(2)

Ajay Kumar et al.: Comprehensive framework for optimization of overlapping pixels in optically butted focal. . .

Fig. 7 Relative MTF in the butting region for four optical system configurations with
(a) f Ny ¼ 0.2f cutoff , (b) f Ny ¼ 0.4f cutoff , (c) f Ny ¼ 0.6f cutoff , and (d) f Ny ¼ 0.8f cutoff .

SNR is ensured in the vignetting region. Hence, it is important to have SNR criterion along
with MTF criterion for optimizing overlap region.
Based on the study of Fuqiang Li et al.,11 where 75% drop in MTF was allowed. xd0 and xr0 are
the solutions for Eq. (16) considering 75% MTF drop for direct and reflected plane, respectively.
The overlap region OMTF ðf u ; f v ; MTFÞ is given in Eq. (17) provided that xd0 < xr0 . If xr0 < xd0
results will lead to under lap, therefore in such cases, MTF shall not be considered as governing
criteria:

OMTF ðf u ; f v ; MTFÞ ¼ jxd0 j þ jxr0 j:

EQ-TARGET;temp:intralink-;e017;116;301 (17)

Figure 7 shows the expected performance based on diffraction limited design and shall be
considered for initial/preliminary assessment only. A detailed analysis of the finalized telescope
configuration should be carried out and achievable MTF should be considered in the framework.
The criteria for MTF will depend on the system design.

3.4 Comprehensive Multicriteria-Based Approach for Optimal Overlapping

Pixel Selection
While designing an optical butting system, all the above discussed parameters should be
considered. Considering the multivariable dependence of overlap requirements in optical butted
focal lane assemblies, a multicriteria approach is developed, as shown in Eq. (18), to derive
optimal overlapping pixels to enable seamless mosaicking of image segments:

total overlap ¼ AUOC þ maxðOMTF ðf u ; f v ; MTFÞ; OSNR ðSc ; N d ; N r ÞÞ:

EQ-TARGET;temp:intralink-;e018;116;123 (18)

The AUOC results from various factors as shown in Eq. (2). This error will lead to uncer-
tainty in defining butting mirror edge position. Reference for SNR and MTF criteria is

Optical Engineering 023103-9 February 2023 • Vol. 62(2)

Ajay Kumar et al.: Comprehensive framework for optimization of overlapping pixels in optically butted focal. . .

determined by butting mirror edge. Hence, AUOC error will be additive to the overlap estimation
arrived using SNR and MTF criterion. The maximum of overlap due to SNR and MTF criterion
ensures overlap region obtained will always enable seamless mosaicking process by ensuring
better image quality in the vignetting region.

4 Case Studies
The proposed framework is test validated through three case studies for different sensor systems
with F∕# 4, 8, and 12 (these will be referred to as designs 1–3, respectively). In all design
configurations, butting mirror is considered to be placed at a 50-mm distance from the focal
plane. All three designs are considered for paraxial and diffraction limited performance.
The sampling frequency of all three systems is 50 lp/mm. All systems use the same focal
plane hence noise performance of the system is the same. Table 1 compares the three systems
with respect to optical butting. The geometric alignment error is assumed to be 0.5 mm and
is identical in all the systems. For ease of understanding, two extreme cases of SNR constraints
are considered.
Considering the approach proposed by Fuqiang et al., the overlap regions for design 1 and
design 2 work out to be ∼7.42 and 1.56 mm, respectively. In case of design 3, it suggests that
optical butting is not be suitable for this design as there exist no pixel pair where the MTF drop
meets the proposed limits.
In case of the proposed scheme, for photon noise limited performance, the overlap
region should be 13, 6.75, and 4.75 mm for designs 1–3, respectively, as discussed in
Sec. 3.2. It is shown in Fig. 8 that in the vignetting region ∼10% drop in SNR is seen if MTF
criterion alone is considered, whereas no drop in SNR is observed considering proposed
scheme. Similarly in design 3, relative MTF in the vignetting region drops to 20% even at
the center field as shown in Fig. 9. The proposed method helps in making design 3 also
an useful configuration by considering SNR criteria, which was not possible with MTF alone
criteria.
As the input signal decreases, the system tends toward electronics noise limited system,
which is the generally the case of high-resolution payloads. The overlap region estimated
considering for designs 1–3 are 7.92, 2.6, and 1.9 mm, respectively.

Table 1 Comparison of the optical butting scheme for designs 1–3.

Parameters Design 1 Design 2 Design 3

F ∕# 4 8 12

Nyquist/cutoff frequency (%) 13.5 27 40.5

The extent of the vignetting region (mm) 12.5 6.25 4.2

AUOC (mm) 0.5

O MTF ðf u ; f v ; MTFÞ (mm) 7.42 1.56 Not possible

Photon-noise limited system

O SNR ðSc ; N d ; N r Þ (mm) 12.5 6.25 4.2

Total overlap (mm) 13 6.75 4.75

Electronics noise limited system

O SNR ðSc ; N d ; N r Þ (mm) 4.2 2.1 1.4

Total overlap (mm) 7.92 2.6 1.9

Optical Engineering 023103-10 February 2023 • Vol. 62(2)

Ajay Kumar et al.: Comprehensive framework for optimization of overlapping pixels in optically butted focal. . .

Fig. 8 Comparison of SNR between Fuiqiang et al. and proposed scheme for design 1.

Fig. 9 Relative MTF in the vignetting region of design 3.

5 Conclusions
This paper describes a comprehensive framework for the optimization of overlapping pixels in
the optically butted FPA. The framework is based on a multicriteria quantitative analytical
approach that accommodates alignment uncertainties errors, MTF, and SNR degradations inher-
ent to the optical butting configuration. As the design and development of optically butted focal
plane assemblies is a very complex process, this framework will provide a new dimension in
ensuring seamless mosaicking of image data from many individual detectors.
This study demonstrates how the framework can be applied to a wide variety of sensor
configurations and how the multicriteria approach helps in determining optimal overlapping
pixels in complex focal plane assemblies at the design phase. Specifically, the implementation
of the proposed comprehensive framework will ensure identifying various influencing factors
and encourage the system engineers to control these factors not only during the design phase but
also during the development phase. To determine optimum overlap pixels, a SNR criterion has
been developed. The SNR criterion reveals that minimum 30% of overlap is required in the
vignetting region. The proposed framework is generic as it incorporates both SNR and MTF
criteria and provides a unified structure for the quantitative analysis of optically butted focal
plane assemblies.

Optical Engineering 023103-11 February 2023 • Vol. 62(2)

Ajay Kumar et al.: Comprehensive framework for optimization of overlapping pixels in optically butted focal. . .

Acknowledgment
The authors would like to thank Shri Soumya. S. Sarkar, deputy director, Sensors Development
Area, for his inspiration and guidance throughout our work. The authors would also like to thank
the Director, Space Applications Centre, ISRO for his valuable suggestions, encouragement, and
support during the entire course of work.

References
1. R. Le Goff et al., “Focal plane AIT sequence: evolution from HRG-Spot 5 to Pleiades HR,”
Proc. SPIE 10567, 105671O (2006).
2. K. Jacobsen et al., “High resolution satellite imaging systems-an overview,” Photogramm.
Fernerkundung Geoinf. 6, 487–496 (2005).
3. G. A. Luppino, J. L. Tonry, and C. W. Stubbs, “CCD mosaics—past, present, and future:
a review,” Proc. SPIE 3355 (1998).
4. J. Xin et al., “Alignment method of optical registration for multi-channel CCD camera,”
Proc. SPIE 10155, 1015536 (2016).
5. A. Sarkar et al., “Development of method for active alignment of multiple time delay and
integration detectors in the optically butted focal plane assembly of high resolution space
borne imaging systems,” J. Appl. Remote Sens. 12(3), 034003 (2018).
6. L. Zhang et al., “Multi-image matching for DSM generation from IKONOS imagery,”
ISPRS J. Photogramm. Remote Sens. 60, 195–211 (2006).
7. E. Baltsavias et al., “Extraction of geospatial information from high spatial resolution
optical satellite sensors,” in ISPRS Tech. Commission IV Symp. “Geospatial Databases for
Sustainable Development”, 27-30 September, Goa, India (2006).
8. J.-L. Lamard et al., “Design of the high resolution optical instrument for the Pleiades HR
Earth observation satellites,” Proc. SPIE 10568, 105680J (2004).
9. D.-H. Lee et al., “Image restoration of calibration and validation for Kompsat-2,” Int. Arch.
Photogramm. Remote Sens. Sp. Inf. Sci. 37, 57–62 (2008).
10. C. Aguerrebere et al., “Fundamental limits in multi-image alignment,” IEEE Trans. Signal
Process. 64(21), 5707–5722 (2016).
11. F. Li et al., “Calculation of overlapping pixels for optical-butting focal plane,” Proc. SPIE
10020, 100200M (2016).
12. J. W. Goodman, Introduction to Fourier Optics, 2nd ed., p. 139, Mcgraw Hill Company Inc.
(1996).

Ajay Kumar Krishnakumar received his BTech degree in physical sciences from the Indian
Institute of Space science and Technology, Thrivanthapuram, Kerala, India. He joined the
Space Applications Centre, Indian Space Research Organization, Ahmedabad, India, in October
2013 and has been working there since then in system design, testing, and characterization of
space-borne imaging systems. His research area of interest is in the design and modeling of
electro-optical imaging sensors and onboard calibration systems.

Nitesh Thapa received his BTech degree in electronics and communication engineering from
Uttarakhand Technical University, Dehradun, Uttarakhand, India. He joined the Space Applications
Centre, Indian Space Research Organization, Ahmedabad, India, in September 2010 and has
been working there since then in system design, testing, and characterization of space-borne
imaging systems. The focus of his research is on the design and development of electro-optical
imaging systems for space missions and the precision calibration of imaging systems.

Narasiha B. Sharma received his PhD in science from the Institute of Science, Nirma
University Ahmedabad, Gujarat, India. He joined the Space Applications Centre, Indian Space
Research Organization, Ahmedabad, India, in July 1997 and has been working there since then
in the integration, testing, and characterization of space-borne imaging systems for various IRS
missions. The focus of his research on in development of systems and methods for performance
optimization and characterization of space-borne electro-optical imaging systems. Currently,
he is a head of the Sensor Systems Division of EO payloads.

Optical Engineering 023103-12 February 2023 • Vol. 62(2)

Ajay Kumar et al.: Comprehensive framework for optimization of overlapping pixels in optically butted focal. . .

Arti Sarkar received her master of technology (MTech) degree in applied optics from the Indian
Institutes of Technology, Delhi, India. She joined the Space Applications Centre, Indian Space
Research Organization, Ahmedabad, India, in December 1992 and has been working there since
then in the design and development of optical telescopes, performance optimization, integration,
testing, and characterization of space-borne imaging systems for various Indian remote sensing
missions. The focus of her research is on the development of methods for telescope alignment
and performance evaluation of space-borne electro-optical imaging systems. Currently, she is
a group director of the Electro Optical System Design and Integration Group.

Optical Engineering 023103-13 February 2023 • Vol. 62(2)