(12) INTERNATIONAL APPLICATION PUBLISHED UNDER THE PATENT COOPERATION TREATY (PCT)

(19) World Intellectual Property Organization

International Bureau
(10) International Publication Number
(43) International Publication Date ι
7 April 2011 (07.04.2011) WO 201 1/039062 Al

(51) International Patent Classification: (72) Inventors; and

G02B 13/06 (2006.01) H04N 3/15 (2006.01) (75) Inventors/Applicants (for US only): DUPARRE,
G02B 3/00 (2006.01) Jacques [DE/DE]; Bertolt-Brecht-StraBe 29, 07745 Jena
(DE). BRUCKNER, Andreas [DE/DE]; Dornburger Str.
(21) International Application Number:
83, 07743 Jena (DE). WIPPERMANN, Frank [DE/DE];
PCT/EP2010/063741
Markt 10, 98617 Meiningen (DE). ZUFFEREY, Jean-
(22) International Filing Date: Christophe [CH/CH]; Rte de la Brftlee 4B, CH-1024
17 September 2010 (17.09.2010) Ecublens (CH). FLOREANO, Dario [CH/CH]; Chemin
de la Richardaz 7, CH-103 1 Mex (CH). FRANCESCHI-
(25) Filing Language: English
NI, Nicolas [FR/FR]; 34bis Ch. J . Aiguier, F- 13009 M ar
(26) Publication Language: English seille (FR). VIOLLET, Stephane [FR/FR]; 133 rue Bre-
teuil, F-13006 Marseille (FR). RUFFIER, Franck
(30) Priority Data: [FR/FR]; 8 1 rue de la Palud, F-13006 Marseille (FR).
09012409.0 30 September 2009 (30.09.2009) EP
(74) Agents: ZIMMERMANN, Tankred et al; P.O. Box
(71) Applicants (for all designated States except US): Fraun- 246, 82043 Pullach (DE).
hofer-Gesellschaft zur Forderung der angewandten
Forschung e.V. [DE/DE]; HansastraBe 27c, 80686 (81) Designated States (unless otherwise indicated, for every
Munchen (DE). Ecole Poly technique Federate de Lau¬ kind of national protection available): AE, AG, AL, AM,
sanne [CH/CH]; Batiment CE 3.316 Station, CH-1015 AO, AT, AU, AZ, BA, BB, BG, BH, BR, BW, BY, BZ,
Lausanne (CH). Universite de la Mediterranee - Aix- CA, CH, CL, CN, CO, CR, CU, CZ, DE, DK, DM, DO,
Marseille II [FR/FR]; 58, Boulevard Charles Livon, DZ, EC, EE, EG, ES, FI, GB, GD, GE, GH, GM, GT,
Jardin du Pharo, F- 13284 Marseille Cedex 07 (FR). HN, HR, HU, ID, IL, IN, IS, JP, KE, KG, KM, KN, KP,
KR, KZ, LA, LC, LK, LR, LS, LT, LU, LY, MA, MD,
ME, MG, MK, MN, MW, MX, MY, MZ, NA, NG, NI,

[Continued on next page]

(54) Title: ARTIFICIAL COMPOUND EYE AND METHOD FOR FABRICATION THEREOF

232 (57) Abstract: A method for fabricating an imaging sy s

-220
FIG 3A tem, the method comprising providing a flexible substrate
(200), a first layer (220) comprising a plurality of mi-
crolenses (232) and a second layer (240) comprising a
+ plurality of image sensors (242). The method further com
prises stacking the first and the second layer (220; 240)
240 onto the flexible substrate (200) by attaching the plurality
FIG 3 B 242
of image sensors (242) to the flexible substrate, such that
\ 7 / / /r// f / //Y//r//r/// / // each of the plurality of microlenses (232) and image sen
sors (242) are aligned to form a plurality of optical chan
nels (300), each optical channel comprising at least one
FIG microlens and at least one associated image sensor, and
mechanically separating the optical channels (300) such
that the separated optical channels remain attached to the
240 flexible substrate (200) to form a mechanically flexible
imaging system.
FIG 3D
E
240

FIG 3 E
 w o 2011/039062 AI III! II II I III II I l l I I ll l Mill II I I
NO, NZ, OM, PE, PG, PH, PL, PT, RO, RS, RU, SC, SD, EE, ES, FI, FR, GB, GR, HR, HU, IE, IS, ΓΓ , LT, LU,
SE, SG, SK, SL, SM, ST, SV, SY, TH, TJ, TM, TN, TR, LV, MC, MK, MT, NL, NO, PL, PT, RO, SE, SI, SK,
TT, TZ, UA, UG, US, UZ, VC, VN, ZA, ZM, ZW. SM, TR), OAPI (BF, BJ, CF, CG, CI, CM, GA, GN, GQ,
GW, ML, MR, NE, SN, TD, TG).
(84) Designated States (unless otherwise indicated, for even-
kind of regional protection available): ARIPO (BW, GH, Published:
GM, KE, LR, LS, MW, MZ, NA, SD, SL, SZ, TZ, UG,
— with international search report (Art. 21(3))
ZM, ZW), Eurasian (AM, AZ, BY, KG, KZ, MD, RU, TJ,
TM), European (AL, AT, BE, BG, CH, CY, CZ, DE, DK,
 Artificial compound eye and method for fabrication thereof

Description

The present invention relates to an imaging system which may be used as an artificial
compound eye and a method for fabrication thereof.

The vertebrate eye (Fig. 10, left) has provided inspiration for the design of conventional
cameras, which consist of a planar image sensor in the focal plane of single-aperture optics
(single or multiple co-axial lenses). Their purpose is to provide a faithful rendering of the
visual world that can be processed by image analysis algorithms for a large variety of pur
poses, especially for object recognition. However, such vision systems require complex
and fast computation in order to extract motion-related information from a sequence of
high-resolution images. The insect compound eye (Fig. 10, centre and right), instead, con¬
sists of a curved array of microlenses, each conveying photons to a separate set of one or
more photoreceptors. Although the compound eye design offers a comparatively lower
resolution than the vertebrate eye, it is very efficient for local and global motion analysis
over a large field-of-view (FOV), making it an excellent sensor for accurate and fast navi-
gation in 3D dynamic environments. Furthermore, compound eyes take several shapes and
curvatures to fit the head and viewing directions of very different types of insects while
offering the same functionality.

Artificial vision systems inspired upon the insect compound eye will be a radically differ-
ent alternative to conventional cameras and will provide more efficient visual abilities for
embedded applications that require motion analysis in low-power and small packages; arti¬
ficial compound eyes could also adapt to different shapes and curvatures to fit the applica¬
tion requirements. However, the design of artificial compound eyes presents several tech¬
nological and scientific challenges because it drastically departs from the design of con-
ventional cameras for what concerns the components, fabrication procedures, packaging,
and visual processing.

Miniaturized cameras are important features for next-generation customer products. Digital
micro-cameras, which are based on miniaturized classical lens designs, are rarely smaller
than 4 4 x 3 mm 3 for a VGA (Video Graphics Array) with approximately 60° field-of-
view. Recent improvements of CMOS (Complementary Metal Oxide Semiconductor) im¬
age sensors could allow further miniaturization, but the shrinking pixel size will lead to
further loss of sensitivity. Furthermore the thickness of conventional imaging optics is
typically in the same order of magnitude as is their focal length. The digital sensor diago¬
nal and the given field-of-view determine the required magnification of the lens. For a
conventional lens this magnification is defined by its focal length, while for compound
eyes it is not. This is why artificial compound eyes could be much thinner than classical
cameras for the same magnification. Furthermore, in compound eyes each channel could
be individually tuned for its specific viewing direction - large fields of view are possible
without major change in the axial arrangement while in classical lens designs large fields
of view require increasingly more lenses in the axial stack ("fish-eye") which increases the
objective thickness dramatically. Consequently a simple scaling of the classical imaging
approach to the desired size has drawbacks. Furthermore, high-resolution images are not
always required. Often the main aim is a compact, robust and cheap vision system per¬
forming only a very specific task, but at extremely high speed, such as motion detection,
distance estimation, or vision-based stabilization. When these tasks are performed with
high resolution conventional cameras, large field-of-view and complex image processing
are often required, which translate into thick optics and power-consuming computing de¬
vices that are often larger than the optoelectronic device itself.

There are different ways to enlarge the field-of-view of an imaging system. One common
approach is to use omni-directional cameras composed of a mirror and of a conventional
camera (see Gluckman, J. and Nayar, S.K. (1998) Ego-Motion and Omnidirectional Cam¬
eras, Int. Conference on Computer Vision, pp 999-1005; Vassallo, R.F., Santos-Victor, J .
and Schneebeli, H.J. (2002) A general approach for egomotion estimation with omnidirec¬
tional images, OMNIVIS'02 Workshop on Omni-directional Vision, pp97-103; Mei, C.
and Rivers, P. (2007) Single view point ominidirectional camera calibration, IEEE Conf.
On Robotics and Automation, pp 3945-3950), resulting in a catadioptric imaging system.
Such a system has two constraints. One of them lies in the fact that the lens and the mirror
have to be precisely aligned in order to have a single centre of projection generating pure
perspective images. The other constraint is the large weight and size occupied by such a
system.

Another approach is to use fish-eye lenses (see Courbon, J., Mezouar, Y., Eck, L. and Mar¬
tinet, P . (2007) A generic fisheye camera model for robotic applications, Int. Conf. on In¬
telligent Robots and Systems, pp 1683-1688) instead of a reflecting surface. The associated
problem is that the viewed scene has large distortions. Hence it is difficult to calibrate the
sensor and the acquired image must be rectified for further visual processing.

Using several cameras looking into different directions offers a further possibility for gen¬
erating omnidirectional view. For example, the Argus Eye (see Baker, E., Ogale, P., Ferm-
Her, C. and Aloimonos, Y. (2004) A spherical eye from multiple cameras (makes better
models of the world), IEEE Robotics and Automation Magazine: Special Issue on Pano¬
ramic Robots 11(4), pp 31-38) uses a network of cameras, with each of them occupying a
different part of the visual sphere. Although the underlying calibration and visual process-
ing of the Argus Eye system is straightforward, it needs high-resolution cameras for the
accurate recovery of motion information and hence has a high computational demand.

A recent approach combined several optical mouse chips to cover a wide field-of-view
(Dahmen, H., Millers, A. and Mallot, H.A. (2009) Insect inspired odometry by optic flow
recorded with optical mouse chips, to appear in Flying Robots and Insects, Springer Ver-
lag; Beyeler, A., Zufferey, J.-C. and Floreano, D . (2009) Vision-based control of near-
obstacle flight. Submitted to the Journal of Autonomous Robots, special issue on Visual
Guidance Systems for Small Unmanned Aerial Vehicles). These imagers have an onchip
signal processor that correlates two consecutive images. The limitation of this approach is
that each mouse sensor can extract only a single displacement between two images instead
of the full pattern of optic flow available over the entire field-of-view. Furthermore the
entire vision system composed of 5 or more mouse sensors with their own optics is rela¬
tively bulky and heavy.

The newly gained spectrum of technological capabilities is now raising the question of
which vision principle and technology is most suitable for the acquisition of the desired
information at the desired level of miniaturization. A fascinating approach is to study how
nature successfully solved similar problems in the case of very small creatures. In particu¬
lar, flying insects are an interesting source of inspiration for miniaturized vision systems
because they have some of the most stringent constraints in size and energy consumption
and their survival is largely dependent on vision-based behaviors.

Insect compound eyes display high temporal resolution, small volumes, and large fields of
view at the cost of comparatively low angular resolution (see Land, M . (1997) Visual Acu-
ity in Insects. Annual Review of Entomology, 42, 147-177). For small invertebrates, such
as flies or honeybees, the compound eyes are a perfect solution for quickly obtaining visual
information about the environment without overloading their brain with complex image
processing. A biological apposition compound eye consists of an array of microlenses on a
curved surface. Each microlens is associated with a small group of photoreceptors in its
focal plane. Apposition compound eyes have mainly evolved in diurnal insects such as
flies (Horridge, G., A . (1977) The compound eye of insects. Sci. Am., 237 pp. 108-120). A
single microlens-photoreceptor unit forms one optical channel and is commonly referred to
as ommatidium (Kirschfeld, K . (1974) The absolute sensitivity of lens and compound eyes.
Z. Naturforsch., 29 pp. 592-596; see also Fig. 10, right). Gradient refractive index cones
and pigments, which form opaque walls between adjacent ommatidia, prevent light from
being focused by a channel's microlens onto an adjacent channel's receptor ("crosstalk").
Natural apposition compound eyes contain from several hundreds (water fly) to tens of
thousands (honeybee or Japanese dragon fly) of these ommatidia packed in non-uniform
hexagonal arrays (Land, M., and Nilsson, D.-E. (2002) Animal Eyes. Oxford animal biol¬
ogy series, Oxford University Press).

Only very recently micro-optics technology enabled the generation of highly precise and
uniform microlens arrays with small lens heights and their alignment to the subsequent
aperture arrays with the high accuracy of photolithography. Very recent explorations in
insect-inspired optical systems resulted in imaging devices that use an array of microlenses
interfaced to a CMOS imager (see Duparre, J. and Wippermann, F. (2006) Micro-optical
artificial compound eyes. Bioinspir. Biomim., 1 R1-R16). The great advantage of this
novel design, which is a planar version of the apposition compound eye of insects, is that it
is ultra-thin, monolithic, comparatively simple and ultra-light weight (Duparre, J.,
Dannberg, P., Schreiber, P., Brauer, A. and Tunnermann, A . (2004) Artificial apposition
compound eye fabricated by micro-optics technology. Applied Optics, 43(22) pp. 4303-
4310; Duparre, J., Dannberg, P., Schreiber, P., Brauer, A. and Tunnermann, A. (2005) Thin
compound eye camera. Applied Optics, 44(15) pp. 2949-2956). In past demonstrations
however, only conventional high-resolution image sensors have been used with planar
compound optics and only a small number of pixels have been extracted from those. This
resulted in a waste of costly CMOS area due to the sub-optimal coupling of arrays of
densely packed small CMOS pixels with arrays of much larger microlenses and in a maxi¬
mum field-of-view of 65°.

A major drawback with planar compound optics developed so far consists in large off-axis
aberrations that occur when imaging under large angles of incidence. This problem was
partially addressed in previous work by individually optimizing the microlens shapes for
their specific angles of incidence forming a so-called "chirped microlens array" (Duparre,
J., Wippermann, F., Dannberg, P . and Reimann, A. (2005) Chirped arrays of refractive
ellipsoidal microlenses for aberration correction under oblique incidence. Opt. Exp., 13(26)
pp. 10539-10551). However, for large angles of incidence on a planar microlens array,
vignetting (reduction of the image brightness or saturation at the periphery compared to the
image centre of each microlens) drastically reduces the sensitivity of the device while opti¬
cal channel crosstalk drastically decreases image contrast. Consequently, the planar com
pound eye had a restricted field-of-view of about 65°, which is by far less than that of natu¬
ral compound eyes, which allow almost spherical vision.
A few groups worldwide are currently working on highly advanced components that could
serve as starting points for future curved artificial compound eyes. Radtke, D., Duparre, J.,
Zeitner, U. and Tunnermann, A . (2007) Laser lithographic fabrication and characterization
of a spherical artificial compound eye. Opt. Exp., 15(6) pp. 3067-3077, used a refractive
microlens array whose master is laser-beam written on a curved substrate and undergo sev¬
eral subsequent replication steps. This technologically challenging scheme is still limited to
large radii of curvature of the base substrate leading to a bulky system (several cubic cm)
due to a limited tilt angle in the laser beam writer (10°). Since there was no curved image
sensor that could be coupled to the micro-optics array, the image was relayed to a planar
image sensor by means of an array of transparent pinholes in an opaque curved layer con¬
centric with the lens array and bulky relay optics. Furthermore, the system suffered from
crosstalk and a poor light sensitivity.

Another approach (Jeong, K.-H., Kim, J. and Lee, L.P. (2006) Biologically Inspired Artifi¬
cial Compound Eyes. Science 312, pp. 557-561) suffered from similar problems. Here,
several replication processes applying a flexible polydimethylsiloxane (PDMS) membrane
and self-guiding effects in the exposed polymers were used to fabricate waveguides resem¬
bling the cone-structure of ommatidia with microlenses on the tip that were arranged on a
spherical surface. However, also here no integration with a curved photoreceptor array was
demonstrated or technologically possible. A more hybrid approach combined fanned out
fibers with microlenses on the tip arranged on a spherical surface with the end-tips on a
planar image sensor (Hornsey, R., Thomas, P., Wong, W., Pepic, S., Yip, K. and Krishna-
samy, R . (2004) Electronic compound eye image sensor: construction and calibration.
SPIE International conference on Sensors and Camera Systems for Scientific, Industrial,
and Digital Photography Applications V, Vol. 5301, pp. 13-24). Although this solution
demonstrated a complete imaging device, achieved a large field-of-view and mechanical
rigidness, it combined spherical compound optics with planar image sensor technology,
which resulted in a complex macroscopic fabrication and assembly technique, a bulky sys-
tern, and no spare room within the volume of the eye. A Japanese team attached single lens
elements to a matrix of organic field-effect transistors attached on a flexible substrate
(Saito, H., Hoshino, K., Matsumoto, K . and Shimoyama, I. (2005) Compound eye shaped
flexible organic image sensor with tunable visual field. IEEE International conference on
Micro Electro Mechanical Systems, pp. 96-99). They claimed that by manually changing
the curvature of the substrate they could obtain a zooming effect. However, the inaccurate
method of joining microlenses with photoreceptors led to poor alignment of the ommatidia,
thus preventing the extraction of useful visual information.
Flexible, organic image sensors are also currently researched (Xu, X., Davanco, M., Qi, X .
and Forrest, S. R . (2008) Direct transfer patterning on three dimensionally deformed sur¬
faces at micrometer resolutions and its application to hemispherical focal plane detector
arrays. Organic Electronics 9, pp. 1122-1 127), but all suffer from the same problems when
need to be combined with microlens arrays with sub-micron precision. An American team
recently reported the procedure for producing a curved optoelectronic camera with the
shape of a human eye (Ko, H . C , Stoykovich, M . P., Song, J., Malyarchuk, V., Choi, W.
M., Yu, C.-J., Geddes, J. B., Xiao, J., Wang, S., Huang, Y. and Rogers, J . A. (2008) A
hemispherical electronic eye camera based on compressible silicon optoelectronics. Na-
ture, 454 pp. 748-753; Kim, D.-H., Song, J., Choi, W . M., Kim, H.-S., Kim, R.-H., Liu, Z.,
Huang, Y. Y., Hwang, K.-C, Zhang, Y.-W. and Rogers, J. A . (2008) Materials and non-
coplanar mesh designs for integrated circuits with linear elastic responses to extreme m e
chanical deformations. PNAS 105 (48) pp. 18675-18680). A key aspect of this device is
the flexible wiring between adjacent silicon photoreceptors arranged on a lattice, which
allows compression of the lattice. However, the resulting concave silicon optoelectronics is
coupled with a single lens, just as in human eyes, and thus sidesteps the alignment and
crosstalk problems that would arise when hypothetically coupled to a curved array of mi-
crolenses to realize a curved artificial compound eye.

The potential use of these technologies for curved compound eyes presents two additional
problems: one problem consists of precisely aligning the curved microlens surface with the
curved imaging surface; the other problem is that the processes used for fabricating the
vision chip allow for only large and simple sub-pixel structures, thus ruling out the realiza
tion of adaptive photoreceptors (such as a VLSI vision chips). Furthermore, although or-
ganic optoelectronics are very promising, the pixel elements have relatively low efficiency
over the visual spectrum, high dark current and short life cycles due to poorly understood
aging processes (Briitting, W. (2005) Physics of Organic Semiconductors. Wiley- VCH).

In summary, what is often claimed as "artificial ommatidia" or "artificial compound eye"

in the literature often lacks either the photoreceptors or the optics because the teams are
specialized in only one of the two fields or because incompatible technologies are used.
Instead, in the case of curved lens or image sensor arrays obtained by applying stress to a
deforaiable PDMS (polydimethylsiloxane) membrane, the resulting device displays poor
contour accuracy and repeatability when stretched, which for artificial compound eyes
would mean misalignment and thus poor correlation between individual microlenses and
photoreceptors that form an ommatidium.
Hence, it is an object of the present invention to provide an improved concept for artificial
compound eyes.

The object is solved by a method for fabricating an imaging system according to claim 1
and a mechanically flexible imaging system according to claim 17.

The technological key to functional and usable curved artificial eyes is a highly accurate
alignment of microlenses and photoreceptors to an array of optical channels or ommatidia
and the positional and rotational correlation within the array on a common flexible sub-
strate. It is the finding of the present invention that the highly accurate alignment can be
achieved when the integration of an optics layer (microlenses) with an optoelectronics
layer (photoreceptors) and with a flexible layer is performed in the planar stage and the
curvature is introduced at a later stage, when the correlation between the three layers of
each ommatidium is fixed.

Embodiments of the present invention provide a method for fabricating an imaging system,
the method comprising providing a flexible substrate, providing a first layer comprising a
plurality of microlenses and providing a second layer comprising a plurality of image sen¬
sors. The first and the second layer are stacked onto the flexible substrate by attaching the
plurality of image sensors to the flexible substrate, such that each of the plurality of mi¬
crolenses and image sensors are aligned to form a plurality of optical channels, each opti¬
cal channel comprising at least one microlens and an associated image sensor. After the
stacking, the optical channels are mechanically separated such that the separated optical
channels remain attached to the flexible substrate to form a mechanically flexible imaging
system.

According to embodiments, during stacking the first and the second layer onto the flexible
substrate, stacking the plurality of microlenses onto the plurality of image sensors is per¬
formed before attaching the plurality of image sensors to the flexible substrate. According
to embodiments the first and second layer are provided as planar layers, respectively, and
stacking the first layer onto the second layer, attaching the stacked first and second layer
onto the flexible substrate and mechanically separating the optical channels all occur in a
planar state. According to various embodiments, mechanically separating the optical chan¬
nels may comprise cutting, sawing or etching the stacked first and second layer to physi-
cally separate neighbouring optical channels.

Hence, an imaging system, which may serve as a basis for an artificial compound eye, may
be obtained by fabricating individual image sensors, e. g., photodiode-pixels, or small
groups thereof on an image sensor wafer. Also, corresponding microlens or aperture arrays
are fabricated on a microlens or aperture wafer, which may be stacked onto the image sen¬
sor wafer by bonding the two together. The resulting wafer stack may be attached to a
flexible substrate, which may be a flexible printed circuit board (PCB), e.g. a dicing tape.
After attaching the stacked wafers to the flexible substrate, the optical channels or om-
matidia ("camera-cubes") are separated by dicing but leaving them on the flexible sub¬
strate. After that, a portion of the flexible substrate which shall form the artificial com¬
pound eye may be cut out. The obtained flexible imaging system may then be bonded to a
convex basis with a desired radius of curvature such that a resulting inter-ommaditial angle
is well matched to an acceptance angle of optical individual channels or, in the case of us¬
ing several image sensors or pixels per microlens, such that the channel is matched to the
sampling angle as well. Finally, the image sensors or pixels, are electrically connected to
read-out electronics for further data processing.

The flexible substrate may be a flexible PCB (e.g. a dicing tape) wiring out electrical con
nections of the image sensors or photoreceptors. Electrical connections from the image
sensors to the flexible PCB can either be obtained by wire-bonding the image sensors
down to electrical interconnections in the flexible PCB, or by connecting the image sensors
via "through silicon interconnects" (TSI, e.g. VIAS) at the backsides of the image sensors.

From this results a mechanically flexible imaging system comprising a flexible substrate
having attached thereto a plurality of image sensors carrying a plurality of microlenses to
form a plurality of optical channels, each optical channel comprising at least one mi¬
crolens and at least one associated image sensor, and wherein at least some of the optical
channels are mechanically separated to form the mechanically flexible imaging system.

Embodiments of the present invention enable the design, prototyping, programming and
validation of fully functional artificial compound eyes, which are composed of microlens
arrays integrated with adaptive image sensors or photoreceptors, e.g., made of analog very-
large-scale-integration circuits (aVLSI), on flexible electronic substrates. The output of the
artificial compound eyes may be processed by vision filters implemented in encapsulated
programmable devices, such as micro-controllers or field programmable gate arrays
(FPGA) for fast extraction of motion-related information.

Compared to conventional cameras, embodiments of the present invention may offer much
larger field-of-view, nearly infinite depth-of-field (no focusing needed), higher sensitivity,
no image blurring and off-axis aberrations, because the distance between the optical sur¬
face and the photoreceptors will be constant over the entire field-of view, and because each
optical channel will work under perpendicular light incidence for its individual viewing
direction. In comparison with classical cameras where focal lengths, spatial resolution and
field-of-view are intimately coupled, imaging systems according to embodiments of the
present invention allow the use of different focal length for the same field-of-view. Fur-
thermore, the curved shape of the resulting artificial compound eyes may offer space
within the convexity for embedding processing units, battery, wireless communication, and
inertial sensors, such as accelerometers and rate gyroscopes, which may be used for mo¬
tion-related computation. Instead, in conventional cameras these components must be
packed separately because the space between the convex lens and the planar image sensor
must be transparent.

Preferred embodiments of the present invention will be explained in more detail with re¬
spect to the accompanying drawings, in which:

Fig. 1 shows a high-level flow chart of a method for fabricating an imaging system
according to an embodiment of the present invention;

Fig. 2a shows an example of a flexible substrate according to an embodiment of the

present invention;

Fig. 2b shows a wafer comprising a plurality of microlenses according to an em¬

bodiment of the present invention;

Fig. 2c shows a wafer comprising a plurality of image sensors according to an em

bodiment of the invention;

Fig. 3a-e depict a method for fabricating an imaging system according to an embodi¬
ment of the present invention;

Fig. 3f-k depict a method for fabricating an imaging system according to another
bodiment of the present invention;

Fig. 4a shows a top-view of an assembled imaging system before mechanically

separating the optical channels according to an embodiment of the present
invention;

Fig. 4b shows a top-view of an imaging system before mechanically separating the

optical channels according to a further embodiment;
Fig. 5a-c show a mechanically flexible imaging systems according to embodiments of
the present invention;

Figs. 6a-c show different wire-bonding concepts according embodiments of the pre¬
sent invention;

Fig. 7a,b show geometric dimensions of a flexible imaging system mounted to a

curved substrate according toan embodiment;

Fig. 8a-d show perspective views of various flexible imaging systems according to
different embodiments;

Fig. 9 shows a section of a fly head compared to a bio-inspired readout architec-

ture for a cylindrical imaging system according to an embodiment; and

Fig. 10 shows a comparison of a vertebrate eye and an insect compound eye.

Fig. 1 shows a high-level flow chart of a method 100 for fabricating an imaging system
according to an embodiment of the present invention.

In a preliminary fabrication step 110, a flexible substrate, a first layer comprising a plural¬
ity of microlenses and a second layer comprising a plurality of image sensors or photore¬
ceptors are provided. In a following fabrication step 120 the first and the second layer are
stacked onto the flexible substrate by attaching the plurality of image sensors to the flexi¬
ble substrate, such that each of the plurality of microlenses and image sensors are aligned
to form a plurality of optical channels (ommatidia), wherein each optical channel (om-
matidium) comprises at least one microlens and at least one associated image sensor. After
step 120 the optical channels are mechanically separated 130 such that the separated opti-
cal channels remain attached to the flexible substrate to form a mechanically flexible imag¬
ing system.

According to embodiments, during stacking 120 the first and the second layer onto the
flexible substrate, stacking the plurality of microlenses onto the plurality of image sensors
is performed before attaching the plurality of image sensors to the flexible substrate. Ac¬
cording to embodiments the first and second layer are provided as planar layers, respec¬
tively, and stacking the first layer onto the second layer, attaching 120 the stacked first
and second layer onto the flexible substrate and mechanically separating 130 the optical
channels all occur in a planar state. According to different embodiments, mechanically
separating 130 the optical channels may comprise cutting, sawing or etching the stacked
first and second layer to physically separate neighboring optical channels.

In step 110 the provided flexible substrate may be a flexible printed circuit board (e.g. a
dicing tape) as ductile mechanical support for the optical channels. In case of the provision
of a flexible printed circuit board, electrical connections for the plurality of image sensors
to associated readout electronics via the flexible PCB are provided. "Flexible" in the pre¬
sent context means a capability of being bent or flexed without injury or damage. An ex-
ample of a flexible substrate is shown in Fig. 2a.

The flexible substrate 200 shown in Fig. 2a is an example of a one-dimensional flexible

PCB with included electronics 202. Fig. 2a (top) shows a backside of the flexible PCB 200
with wiring 204. The bottom image shows a sensing side of the flexible PCB 200 with in-
dicated placement 210 of 16 one-dimensionally arranged image sensors. The flexible PCB
200 acts as mechanical support for the artificial ommatidia and as electrical connectivity to
readout electronics. The flexible PCB 200 may be very thin, e.g. down to 100 microns,
with 50-100 micron wide electrical lines 204, which may act as connective pads for a wire
bonding process.

Of course, other flexible substrates may be used in other embodiments of the present in¬
vention. In case electrical interconnections are not required within the flexible substrate
(which will be discussed later), the flexible substrate may also be a ductile or bendable
tape, in particular a dicing tape for the (normally temporary) fixing of wafers.

The step 110 of providing the first layer with the plurality of microlenses comprises pro¬
viding a wafer 220 carrying a microlens-array 230 comprising the plurality of microlenses
232, which is exemplarily shown in Fig. 2b.

The first layer 220 carrying the plurality of microlenses 232, i.e., the compound eye optics,
may be designed and/or fabricated by lithographic methods on wafer-scale to achieve a
thickness d in the order of 100 microns. Parameters, such as geometry of the components,
number of ommatidia per patch, optical properties of each ommatidium (field-of-view,
resolution, sensitivity, overall size constraints), architecture of the underlying vision chip
(photoreceptor size and layout, electrical circuitry to be built-in bonding pads), and the
way the photoreceptors are electrically interconnected, may serve as input for the layout of
the microlens array 230 and underlying aperture arrays (not shown) for suppression of op¬
tical channel crosstalk. Several arrangements for one or several photoreceptors per optical
or visual channel are possible. The first layer 220 carrying the plurality of microlenses 232
may be a wafer comprising a glass or polymer substrate for the microlenses 232. The first
layer 220 may be combined with aperture arrays and polymer spacer layers to micro-optic
sandwiches of e.g. glass substrates, aperture arrays and polymer spacer layers.

Fig. 2c shows a schematic side-view of the second layer 240 comprising the plurality of
image sensors or photoreceptors 242. The second layer 240 may be fitted underneath the
first layer 220 carrying the microlens array 230. However, vision chips comprising the
image sensors 242 cannot be conventional imagers for several reasons. First, they need to
provide space around each photoreceptor 242 for the separation step 130 that will occur
once the first layer 220 carrying the microlens array 230 has been bonded, e.g. by gluing,
on top of the image sensor wafer 240. The fact that each photoreceptor 242 may be singu-
lated also means that each of them may need connecting pads for wire bonding to the un¬
derlying flexible substrate 200 or directly to neighboring optical channels. In addition, pas-
sive circuitry 244, e.g. front end electronics like a read-out electronic, may be added next
to each photoreceptor 242. External bond pads 246 serve for connecting the second layer
240, i.e. the optoelectronic photoreceptors 242, to further (e.g. back end) read-out circuitry
(not shown). The image sensor chips may extend beyond at least a part of the periphery of
the optic to allow access to the bond pads on the image censor chip also after having ap-
plied the optic to the image sensor chip.

According to an embodiment, a VLSI technology may be used to provide very low-power

image sensor circuits by relying on sub-threshold methods (see Liu, S., Kramer, J., In-
diveri, G., Delbriick, T., and Douglas, R . (2003) Analog VLSI: Circuits and Principles.
MIT Press, Cambridge, MA). Examples of additional functionalities include the enhance¬
ment of the dynamic range by using a self-adaptive circuit (see Delbriick, T., Mead, C.
(1995) Analog VLSI phototransduction by continuous-time, adaptive, logarithmic photore¬
ceptor circuits. IEEE Computer Society Press, pp. 139-161) that offers a huge range of
operation in terms of light intensity: they can operate in any light condition ranging from
moonlight to full sun ( 1 lux to 105 lux) without any external tuning while consuming very
little power (typically in the order of 50 'ν ) . Such an adaptive photoreceptor 242 is inter¬
esting not only for DC operation, but also for AC operations because it is designed in such
a way that a sudden change in light level is transformed into an impulse that gets back to
the initial resting level with a time constant that depends on the light change amplitude.

Now that components for fabricating the mechanically flexible imaging system, i.e., the
flexible substrate 200, the first layer 220 of microlenses and second layer 240 of image
sensors, have been explained, Figs. 3a-e show a method for fabricating the flexible imag-
ing system according to an embodiment of the present invention. As can be seen, when
stacking the first and the second layers 220, 240 shown in Figs. 3a and 3b onto the me¬
chanically flexible substrate 200 (see Fig. 3c), a stacking of the plurality of microlenses
220 onto the plurality of image sensors 240 is performed before attaching the second layer
240 comprising the plurality of image sensors 242 to the flexible substrate 200 (see Fig.
3d). The first and the second layers 220, 240 are stacked onto each other such that each of
the plurality of microlenses 232 and image sensors 242 are aligned to form a plurality of
optical channels 300, wherein each optical channel 300 comprises at least one microlens
232 and at least one associated image sensor 242. In order to obtain a mechanically flexible
imaging system the optical channels 300 are mechanically separated (see Fig. 3e), such
that the separated optical channels 300 remain attached to the flexible substrate 200. The
mechanical separation of the optical channels 300 may comprise cutting, sawing, etching,
dicing, or laser dicing the stacked first and second layers 220, 240 to physically separate
neighboring optical channels 300. As it can be seen, the flexible substrate 200 is not cut or
mechanically separated, because it acts as mechanical support for the optical channels 300.

Figs. 3f-k show a method for fabricating the flexible imaging system according to another
embodiment of the present invention. The first and the second layers 220, 240 shown in
Figs. 3f and 3g correspond substantially to those of Figs. 3a and 3b, except that in the im-
age sensor layer 240 the front end electronics (e.g. read-out electronics) are integrated such
that each image sensor has associated therewith a front-end electronic circuit 244. The first
and the second layers 220, 240 shown in Figs. 3f and 3g are combined (see Fig. 3h) and
attached onto the mechanically flexible substrate 200. While Figs. 3f-h show side views of
the first and the second layers 220, 240, Figs. 3i-j show top views of the separated layers.
Fig. 3i shows the arrangement of a single image sensor 242 per column while Fig. 3j shows
an embodiment comprising a plurality of image sensors 242 per column. As can be seen
from Figs. 3i-j, the front-end electronics 242 are arranged at the extremity of the layer and
the pads 246 are arranged in the respective column, i.e. along a direction perpendicular to
the bending direction. Fig. 3k is again a side view showing the mechanically flexible imag-
ing system having the optical channels 300 which are mechanically separated in a manner
as described above.

Thus, in the embodiment described with reference to Figs. 3f-k, the front-end electronics
and the bond pads are integrated and placed at the extremity of each column, so that other
than in the embodiments shown in Figs. 2c and 3a-e there is no need to arrange the read¬
out electronics and pads at the periphery of the layer 240. In this embodiment the read-out
electronics and pads are simply integrated and arranged at the tip of each photosensor array
or each column of ommatidia. There is no need to place the electronics at a back side but
just near the pixels of each columns' extremity. This arrangement is advantageous as it
avoids the wire-bonding between each column so that bending the flexible substrate will
not damage the wire connections.

According to embodiments, a precise combination of microlens- and photoreceptor arrays

onto the mechanically flexible substrate 200 including electrical wiring and encapsulation
may be achieved due to the fact that all assembly steps occur in a planar state, which a l
lows high precision, while the final optical channel or ommatidia patch will be completely
mechanically flexible. I.e., providing the first and second layer 220, 240 comprises pro-
viding planar layers, respectively, wherein stacking the first layer 220 onto the second
layer 240, attaching the stacked first and second layer onto the flexible substrate 200)and
mechanically separating the optical channels 300 occur in a planar state. To achieve the
alignment of the optical dies (first layer 220) with the photoreceptor dies (second layer
240) a matching of alignment marks in both structures may be used. Through special chip
bonding a sub-micron precise lateral and rotational alignment of the image sensors 242
with respect to their corresponding microlenses 232 is possible. During chip bonding ex¬
tremely short focal lengths of the microlenses 232 has to be dealt with. An axial or vertical
distance between the optics layer 220 and the image sensors 242 of the ommatidia 300
could be determined by the amount and viscosity of an applied UV-glue and/or by active
setting via manipulation stages. The optical channels 300 or columns thereof may be singu-
lated from each other by wafer dicing only once the interconnection between the lenslets
232 and the vision chip 240 is permanently fixed and the lens-vision-chip-sandwich has
been applied to the flexible PCB 200, which at the same time serves as a dicing tape. The
electrical connection from the image sensors 242 to the flexible PCB 200 or directly be-
tween the optical channels or ommatidia 300 themselves may be achieved by wire bond¬
ing.

The optical channels or ommatidia 300 are the building blocks of curved compound eyes.
Just like in nature, the proposed mechanically flexible ommatidia 300 will have to display
the same structure across the entire surface of the compound eye, no matter what shape and
curvature the eye takes. According to embodiments of the present invention a density of
optical channels or ommatidia 300 is 2 - 10 ommatidia per mm 2 . As has been explained
before, the ommatidia 300 are made of a first layer 220 of custom made microlenses 232
aligned on top of a second layer 240 of a custom-made aVLSI vision chip and further
bonded to a flexible PCB 200. The first two layers 220 and 240 of the obtained sandwich
are then diced in order to mechanically separate the visual channels 300 (ommatidia singu¬
lation) and to enable the bending of the mechanically flexible substrate 200.
A schematic top-view of the obtained mechanically flexible imaging system is given in
Fig. 4a.

It can be seen, that the top layers 220, 240 are diced, as indicated by the dashed vertical
lines, in order to mechanically separate neighboring rows or columns of optical channels
300. Thereby the sensor 240 is not affected by the separation process. The microlenses 232
may be chirped for correction of off-axis aberrations. In order to obtain a small vision field
when bending the flexible imaging system cylindrically, the microlenses 232 and the im¬
age sensors 242 may be displaced, like schematically shown in Fig. 4a, right. Here, dis-
placed means that the microlenses 232 and their associated image sensors 242 do not lie on
a common vertical axis. After the column-wise separation of the optical channels 300, the
sensor 240 may be mounted to a bent surface, for example, a cylindrical surface, as indi¬
cated by the arrow 4 10 in Fig. 4a.

As is schematically shown in Fig. 4b, also a plurality of image sensors 242 may be associ¬
ated to a single microlens 232, such that an optical channel 300 comprising the microlens
232 and the plurality of image sensors 242 serves for getting a column image perpendicular
to the curved spreading of the optical-channel-columns. The curved or cylindrical spread¬
ing of the flexible imaging system is shown in Figs. 5a, b.

Optical viewing directions of the individual separated optical channels 300 are tunable
according to a radius of curvature of a supporting structure (not shown in Fig. 5a). This
allows to obtain a desired ratio between viewing angle and field-of-view of the individual
optical channels 300, repeatable across the whole compound eye.

Fig. 5b shows a side-view of a flexible imaging system comprising mechanically separated

optical channels 300 attached to the flexible substrate 200, wherein the flexible substrate
200 is glued to a semi-cylindrical curved basis 500. The curved convex basis 500 thereby
provides enough space for additional electronics, e.g., optic flow processing units (FPGA
or DSP), antenna or battery circuits.

Due to the physical separation of the optical channels 300 on the PCB 200, there are cavi¬
ties or grooves 510 between separated optical channels 300. Screening of the ommatidia or
optical channel side-walls with opaque curable material in order to prevent optical
crosstalk between neighboring ommatidia or optical channels 300 is possible. The screen¬
ing and encapsulation with opaque curable could also serve as stabilization and encapsulat¬
ing for a curved arrangement of the ommatidia. Hence, according to embodiments, the
cavities 510 between the separated optical channels 300 may be filled with a sealing com-
pound material 520. The sealing compound material 520 may be a optically isolating seal¬
ing compound material for isolating neighboring optical channels 300. After filling the
sealing compound material 520 into the cavities 510, the flexible substrate 200 carrying the
separated optical channels 300 may also be removed from the curved basis 500. Due to the
filled cavities 510 the curved shape of the imaging system will remain even after removing
it from the curved basis 500.

Fig. 5c shows a perspective view of a semi-cylindrical curved compound eye imaging sys¬
tem having separate rows or columns 530 of optical channels 300. Each optical channel
comprises a microlens 232 and at least one image sensor 242 on a image sensor die 540,
the image sensor die 540 bonded to a flexible PCB 200 by bond wires and bond pads 246.

Different wire-bonding schemes will now be explained with reference to Figs. 6a, 6b.

Fig. 6a shows an example of front side wire-bonding an optical channel 300 to the flexible
substrate or PCB 200. Contact pads 602 of an image sensor die 604 are connected to the
flexible PCB 200 by bond-wires 606. Further, passive electronics 244 associated to the
image sensor 242, such as, e.g., column amplifiers, may be foreseen on the flexible PCB
200.

Fig. 6b shows an example of wire-bonding an optical channel 300 to the flexible substrate
or PCB 200 with TSV (through silicon via). The image sensor 242 (or image sensor array)
is connected to the flexible PCB 200 by means of through-hole silicon vias 612 and solder
bumps 614. Thereby, the electrical connections 612 may be provided by correspondingly
etching the substrate (e.g. silicon) of the image sensor die 604.

In both examples according to Fig. 6a, b electrical connections provided by means of the
flexible substrate 200 may be realized e.g. by printing contact bands and/or electrical con¬
nections on the flexible substrate 200 e.g. with printable conductive ink.

A further alternative for connecting the image sensors with their periphery is shown in Fig.
6c. Here, pixels or image sensors 242 are connected with their respective periphery 244
and external bond pads 620 by bonding wires 606 between individual silicon pieces. There
is no down-routing to the flexible substrate 200 and, hence, in this case the flexible sub-
strate 200 does not need to have electrical connection properties.
After having explained embodiments of the fabrication method and the flexible imaging
system itself, a design example of the flexible imaging system according to an embodiment
of the present invention is now given referring to Figs. 7a, b .

A semi-cylindrical curved surface or basis 500 with radius r = 5 mm may be considered.

The resulting arc length is 1 = π r « 15 mm (half a cylinder). Microlenses 232 with a di¬
ameter D = 150 µη and a focal length of f = 400 µ η may be used. An angular horizontal
spacing ∆φι, may be 6°. Resulting therefrom are 180°/6° = 30 lenses to cover the entire
semi-cylinder 500. In this case an acceptance angle ∆ρ of also 6 ° is needed. A pixel or
image sensor 242 of 40 µη x 40 µιη and a focal length of about 400 µ η may be used. The
interval (free space) between each lens 232 will be (15 - 30*0.15)/ 30 = 350 µ η . Thus, the
pitch p will be 350 µηι + 150 µ = 500 µηι . This pitch value should be large enough to let
the resulting optical channel 300 be diced without any damage. A vertical interommatidial
angel ∆φν may be equal to ∆φι, for a field-of-view of 60°. The main limit lies in the ability
to shift the visual axis. This means that each column of the imaging system will be com¬
posed of ten optical channels 300 (ommatidia), like shown in Fig. 7b. Even if such a imag¬
ing system or compound eye may have a rather coarse resolution, it is well suited and
dedicated to optic flow measurement. In any case, it can be considered as a classical
imager yielding an image that can be useful for the human eye.

Flexible imaging systems according to embodiments of the present invention may be the
building blocks for a realization of fully functional curved compound eyes. After the singu¬
lation of the ommatidia 300, or patches of them (see Fig. 8a), on a flexible PCB, which at
the same time serves a dicing tape, the optical viewing directions are tuneable according to
the radius r of curvature of a supporting structure 500. This will allow to obtain a desired
ratio between viewing angle and field-of-view of the individual ommatidia, repeatable
across the whole compound eye.

With embodiments of the present invention various types of curved compound eyes are
possible, e.g., cylindrical compound eyes (see Fig. 8b), active or actuated compound eyes
wherein a cylindrical compound eye is endowed with scanning movements, spherical eyes
(see Fig. 8c) and tape-like compound eyes (see Fig. 8d).

The cylindrical compound eye shown in Fig. 8b may offer a very large field-of-view in the
horizontal plane (up to 360°). The vertical columns of ommatidia - when not diced - may
be designed to display a slight difference in pitch between the photoreceptor and mi¬
crolenses (see Fig. 4a, right), such that the vertical field-of-view could span up to 60°, even
without adding vertical curvature. The cavity of a cylindrical compound eye may offer
space for encasing processing units and small actuators. A great advantage may be seen in
the free space for components in the eye cavity, which may result in very compact and
self-sufficient sensing devices. In the imaging system according to Fig. 8b, the plurality of
optical channels 300 is overall arranged in a two-dimensional array of optical channels,
wherein neighbouring rows 530 of optical channels are mechanically separated.

The columnar arrangement of ommatidia 300 seems to be the easiest way of providing a
2D field of view using only ID flexibility at the level of the substrate 200. Since a detec¬
tion of fast motion-related events shall be detected, bio-inspired electronic architectures
(Fig. 9, right) featuring distributed analog-to-digital converters (ADC) and programmable
processing units may be used. This may enable fast visual signal acquisition and local
processing before sending the information further to a central processor for global visual
processing. For instance, a micro-controller may be integrated at the column level to proc¬
ess locally the optic flow. In the case where a column is composed of more than 1 pixel
(2D retina), a local microcontroller could process the optic flow measured by each vertical
ommatidia. In any case, a programmable processing unit (microcontroller or FPGA) may
be added to process all the incoming signals and compute global information.

The active, or actuated compound eye may be seen as an cylindrical eye augmented with a
device providing a very small minimum visibile (i.e., the smallest area distinctly perceiv¬
able by the eye) and an outstanding hyper acuity (Viollet and Franceschini 2001, Viollet
and Franceschini 2005). These two properties, rarely encountered in any artificial visual
sensor, will rely on the ability of the physical sensor to generate controlled vibrations. In
the context of actuator technology for micro-actuation systems, at least two technologies
may be considered: piezo and artificial muscles, or shape memory alloy (SMA). Piezoelec¬
tric actuators are widely used due to their high speed, high resolution, good mechanical
durability, high output force and low power consumption (Tao et al., 2007). For example,
piezo bender actuators have been widely used in academic studies and industrial applica¬
tions, e.g., as piezoelectric switch, valve, pump, ultrasonic motor, printer head, piezoelec-
trie fan, piezoelectric tube scanner and quick focusing lenses (e.g, Yoo et al., 2000). The
success of piezo technology relies on its high bandwidth (e.g, about 500 Hz). This means
that piezo actuators can be controlled in open loop, leading to a drastic simplification of the
control electronics. However, the necessary high voltage electronics required for actuating
piezo actuators is often complex and bulky. These requirements also limit the use of pie-
zoactuators in autonomous micro-robots (Stelz et al., 2006). However, novel architectures
based on boost converter technology open ways to overcome the limitations imposed by
the bulky size of power piezo driver. Even if piezo technology plays a major role in micro-
actuating, recent developments in shape memory alloys may lead to SMA actuators for
implementing the micro-scanning of the imaging system. E.g., Biometal fiber developed
by Toki Inc. are currently used in a smart 0.8 g servo actuator including a very small elec¬
tronic driver. This kind of actuators is now used in many research fields.

A possible realization of a spherical compound eye (Fig. 8c) consists of arranging several
ommatidia patches around a sphere, thus enabling a truly omnidirectional field of view and
providing extra free space in the eye cavity for additional components, such as accelerome-
ters and rate gyros, which are necessary for some types of motion-related analysis. For
example, the rotational component of optic flow, which does not correlate with distance to
surrounding objects, can be discounted if the rotational components of the eyes are avail¬
able from 3-axis rate gyroscopes (Rodriguez, A., Andersen, E., Bradley, J. and Taylor, C.
(2007) Wind Estimation Using an Optical Flow Sensor on a Miniature Air Vehicle. AIAA
Conference on Guidance, Navigation, and Control; Zufferey, J.-C. and Floreano, D. (2006)
Fly-inspired Visual Steering of an Ultralight Indoor Aircraft. IEEE Transactions on Robot-
ics, 22(1) pp. 137-146; Zufferey, J.-C, Beyeler, A . and Floreano, D . (2009) Optic Flow to
Steer and Avoid Collisions in 3D. In Flying Insects and Robots, Berlin : Springer- Verlag.
In press).

A lightweight artificial spherical compound eye can cover a truly omnidirectional field of
view in order to emulate the visual capabilities of insects. Among various possibilities, a
series of circular PCBs may be used or several lines of ommatidia may be arranged directly
onto a spherical surface (Fig. 8c). The former option may facilitate the interconnection of
the ommatidia, whereas the latter could be easier to integrate onto the surface of a vehicle
or a robot. Since it is well known that the inertial sense is of high importance to interpret
visual information such as optic flow, the free space available within the spherical com¬
pound eye may be exploited to add a set of inertial sensors: 3-axes gyroscopes and 3-axes
accelerometers.

Tapes of elongated and flexible compound eyes (Fig. 8d) may consist of bendable printed
circuits embroiled with multiple ommatidia, to be attached to various surfaces or sewed
into clothes. In the imaging system according to Fig. 8d, the plurality of optical channels
300 is arranged in a one-dimensional array of optical channels, wherein neighbouring opti¬
cal channels are mechanically separated. Adding other components may also be consid¬
ered, such as wireless modules and batteries, which could be soldered on the tape, for mak-
ing it energetically self-sufficient and capable of communicating motion-related informa¬
tion to other devices. The viewing directions of the ommatidia is determined by the under¬
lying surface and not fixed at fabrication time.
Elongated vision-tapes can be attached onto various flat and curved surfaces or sewed into
clothes in order to perceive motion-related information in divergent directions. Unlike
standard cameras, which occupy a volume because of their single lens architecture, this
vision-tape may cover a significantly wider surface of several centimeters while being only
a few micrometers thick. This skin-like property may allow covering widely divergent
viewing directions (depending on the underlying surface) while remaining almost unno-
ticeable from a mechanical point of view. An interesting aspect with this special type of
imaging system is that we cannot assume a priori knowledge of the repartition of the view¬
ing directions of the ommatidia since the support on which the tape will be glued could be
of any shape or even moving and flexing while the compound eye is operational (e.g.,
when sewn into clothes).

Two approaches may be considered to actually produce such tapes. The first one would
consist of replicating patches of ommatidia produced according to embodiments of the
present invention and solder them on top of a second flexible PCB of larger size. The sec¬
ond approach would be to pick and place each single ommatidium from its production sub¬
strate onto the longer flexible PCB. The first approach is simpler in terms of assembly pro¬
cedure, but the tape may end up not being homogeneously populated with ommatidia and
the thickness may be greater due to the superposition of two layers of flex PCB in some
places. The second approach instead may involve dissolving the glue that has been used in
the ommatidia production process (once they have been diced apart from each other), care¬
fully pick them one by one and gluing them onto the larger PCB before they can be wire-
bonded and encapsulated. Although time consuming, this latter approach could allow more
freedom in the way the ommatidia can be spread across the tape, which would represent an
advantage. A final form of the vision-tape should include thin batteries and a means of
sending out motion-related information via wired or wireless communication.

Another important aspect of fully functional compound eyes is the design of the readout
circuitry and built-in micro-processing units. Taking inspiration from the fly's visual sys-
tern (Fig. 9, left), which is composed of a series of neuropils responsible for successive
visual processing tasks (local pre-processing -> local motion detection -> global interpreta¬
tion of motion fields), the use of a network of interconnected microcontrollers (Fig. 9,
right) may be considered that may be directly soldered onto the flexible substrate of the
ommatidia patches. Similarly to the lamina and medulla in the fly brain, the local micro-
controllers can be used to pre-process local visual information in order to remove noise,
adapt gains, and estimate optic flow among adjacent ommatidia. Similarly to the lobula in
the brain of the fly, a main processing unit could then integrate the local information proc¬
essed by the distributed microcontrollers into higher level signals such as global optic-flow
fields, distance estimates or ego-motion. All the processing units (e.g. low-power micro¬
controllers or FPGA) may be fully programmable.

To summarize, embodiments of the present invention provide a compound eye imaging

system which may be applied to a curved basis with a highly-precise relation between mi¬
crolenses and photoreceptors which leads to a highly precisely determined sampling angle.
Microlenses and photoreceptors make up optical channels, wherein a flexible substrate
serves as a mechanical connection between neighboring optical channels. Additionally, an
opaque sealing compound may be used for optical channel separation.

According to embodiments the optics comprising the microlenses as well as the image sen¬
sors are provided in planar stacking technique. Both layers are combined in planar condi¬
tion by waferscale or chipscale mounting. Mounting the microlens-layer and the image-
sensor-layer in the planar domain and using established methods allows for a highly pre-
cise relation between microlenses and photoreceptors in the sub-micron area. The bonded
components are then attached to the flexible substrate which may, at the same time, serve
as a dicing tape. In a next step, the optical channels resulting from microlenses and associ¬
ated image-sensors are singulated by, e.g., sewing the microlens-photosensor-sandwich,
however, without "injuring" the flexible substrate. The resulting imaging system gets
flexible or bendable, whereas the association microlens-photosensor within the optical
channels remains fixed. The singulation may be performed via sawing, laser-cutting, etch¬
ing, etc. The obtained flexible imaging device may be bent and/or mounted to a curved
basis. Cavities between the separated optical channels may be filled with a hardening seal¬
ing compound material, which may be preferably a black material for optically isolating
the optical channels. The flexible imaging device may be fixed to the curved basis perma¬
nently, or may be again separated from it after "freezing" the curvature. Alternatively, a
completely flexible or actuated form may be envisaged, like it has been explained above.
In this case however, the current deformation condition has to be known in order to recog¬
nize an object from the image.

Although several aspects of the present invention have been described with respect to an
imaging system it is to be understood that those aspects also serve as a description of a
related method for fabricating the imaging system, such that a block or an element of the
imaging system may also interpreted as a related step or a feature of a step of the method
for fabricating. Also, aspects that have been described in connection with or as a step of a
method may also be interpreted as a description of a respective block, or a detail, or a fea¬
ture of a respective imaging system.
Embodiments of the invention described that the image sensors are connected to the flexi¬
ble substrate by wire-bonding. The invention is not limited to such a technique, rather other
techniques for connecting the image sensors to the flexible substrate may be used, e.g. by
trough silicon interconnection (TS-VIAS), gold bands, printed conductive traces, conduc-
tive ink. Also other know techniques may be used.
 Claims

A method (100) for fabricating a mechanically flexible imaging system, the method
comprising:

providing ( 1 10) a flexible substrate (200);

providing ( 1 10) a first layer (220) comprising a plurality of microlenses (232);

providing ( 1 10) a second layer (240) comprising a plurality of image sensors (242);

stacking (120) the first and the second layer (220; 240) onto the flexible substrate
(200) such that each of the plurality of microlenses (232) and image sensors (242)
are aligned to form a plurality of optical channels (300) , each optical channel com¬
prising at least one microlens and at least one associated image sensor, wherein the
first and the second layer (220; 240) are attached onto the flexible substrate (200)
such that the second layer (240) comprising the plurality of image sensors (242) is
attached to the flexible substrate; and

mechanically separating (130) the optical channels (300) such that the separated
optical channels remain attached to the flexible substrate (200) to form a mechani¬
cally flexible imaging system.

The method according to claim , wherein providing the flexible substrate (200)
comprises providing a flexible printed circuit board as mechanical support for
the optical channels (300).

The method according to claim 2, wherein providing the flexible printed circuit
board (200) comprises providing electrical connection from the plurality of im¬
age sensors (242) to associated readout electronics via the flexible printed circuit
board.

The method according to one of the preceding claims, wherein the plurality of
image sensors (242) are provided using an analogue very large scale integration
technology.
5. The method according to one of the preceding claims, wherein providing the p lu
rality of image sensors (242) comprises providing connection pads (246) for bond¬
ing each of the plurality of image sensors to the flexible substrate.

6. The method according to one of the preceding claims, wherein, during stacking
the first and the second layer onto the flexible substrate, stacking the plurality of
microlenses onto the plurality of image sensors is performed before attaching the
plurality of image sensors to the flexible substrate.

7. The method according to one of the preceding claims, wherein providing the first
and second layer (220; 240) comprises providing planar layers, respectively, and
wherein stacking the first layer (220) onto the second layer (240), attaching the
stacked first and second layer onto the flexible substrate (200) and mechanically
separating the optical channels (300) occur in a planar state.

8. The method according to one of the preceding claims, wherein mechanically

separating the optical channels (300) comprises cutting, sawing, etching, dicing,
or laser dicing the stacked first and second layer to physically separate
neighbouring optical channels (300).

9. The method according to one of the preceding claims, wherein stacking the plu-
rality of microlenses (232) onto the plurality of image sensors (242) comprises
matching previously provided alignment marks on the first and/or second layer
allowing an exact alignment of microlenses with associated image sensors.

10. The method according to one of the preceding claims, wherein stacking the plu-
rality of microlenses (232) onto the plurality of image sensors (242) comprises
gluing the first layer (220) onto the second layer (240).

11. The method according to one of the preceding claims, wherein attaching the plu¬
rality of image sensors (242) to the flexible substrate (200) comprises bonding of
electrical connections of the image sensors with electrical connections of the
flexible substrate.

12. The method according to one of the preceding claims, wherein the method fur¬
ther comprises, after stacking the plurality of microlenses (232) onto the plurality
of image sensors (242) and attaching the plurality of image sensors to the flexi-
ble substrate (200), mounting the flexible substrate carrying the separated optical
channels (300) to a curved surface (500).

The method according to claim 12, wherein, after mounting the flexible substrate
(200) to the curved surface (500), cavities (510) between the separated optical
channels (300) are filled with a sealing compound material (520).
 A . CLASSIFICATION OF SUBJECT MATTER
INV. G02B13/06
ADD. G02B3/00 H04N3/15

According to International Patent Classification (IPC) or to both national classification and IPC

B. FIELDS SEARCHED
Minimum documentation searched (classification system followed by classification symbols)
G02B H04N H01L

Documentation searched other than minimum documentation to the extent that such documents are included in the fields searched

Electronic data base consulted during the international search (name of data base and, where practical, search terms used)

EPp-Internal , WPI Data

C. DOCUMENTS CONSIDERED TO BE RELEVANT

Category* Citation of document, with indication, where appropriate, of the relevant passages Relevant to claim No.

A JP 2003 283932 A (JAPAN SCIENCE & TECH 1-13

CORP) 3 October 2003 (2003-10-03)
* abstract; figures 1-4

VOLKEL R ET AL: "Miniaturized imaging 1-13

systems "
MICROELECTRONIC ENGINEERING, ELSEVIER
PUBLISHERS BV., AMSTERDAM, NL,
vol. 67-68, 1 June 2003 (2003-06-01),
pages 461-472, XP004428906
ISSN: 0167-9317
paragraph [03.4]; figures 3,5

EP 2 088 572 A2 (PALO ALTO RES CT INC 1-13

[US]) 12 August 2009 (2009-08-12)
paragraphs [0016] - [0021], [0031],
[0041] - [0046]; figures 1-3,6

□ Further documents are listed in the continuation of Box C. See patent family annex.

* Special categories of cited documents :

"T" later document published after the international filing date
or priority date and not in conflict with the application but
"A" document defining the general state of the art which is not cited to understand the principle or theory underlying the
considered to be of particular relevance invention
"E" earlier document but published on or after the international "X" document of particular relevance; the claimed invention
filing date cannot be considered novel or cannot be considered to
"L" document which may throw doubts on priority claim(s) or involve an inventive step when the document is taken alone
which is cited to establish the publication date of another "Y" document of particular relevance; the claimed invention
citation or other special reason (as specified) cannot be considered to involve an inventive step when the
"O" document referring to an oral disclosure, use, exhibition or document is combined with one or more other such docu¬
other means ments, such combination being obvious to a person skilled
in the art.
"P" document published prior to the international filing date but
later than the priority date claimed "&" document member of the same patent family

Date of the actual completion of the international search Date of mailing of the international search report

2 1 October 2010 02/11/2010

Name and mailing address of the ISA/ Authorized officer
European Patent Office, P.B. 5818 Patentlaan 2
NL - 2280 HV Rijswijk
Tel. (+31-70) 340-2040,
Fax: (+31-70) 340-3016 Casse, Martin

Form PCT/ISA/210 (second sheet) (April 2005)

 Patent document Publication Patent family Publication
cited in search report date member(s) date

JP 2003283932 03-10-2003 JP 4012752 B2 21-11-2007

EP 2088572 A2 12-08-2009 JP 2009177179 A 06-08-2009

US 2009184954 A l 23-07-2009

Form PCT/ ISA 2 (patent family annex) (April 2005)