IRJET- Image based Approach for Indian Fake Note Detection by Dark Channe...
Novel Panoramic Camera Design
1. Novel design for Panoramic Camera
Dr Monika Jain Mr Dhiraj kr Sati
Professor Vice-President
Dept of Electronics & Instrumentation Engineering DPH Software Services Pvt Ltd.
Galgotias College of Engineering & Technology
Greater Noida
Nehru Place,
New Delhi
monika_jain24@rediffmail.com dksati@gmail.com
Ms Cuckoo Sati Mr Devarshi Bajpai
B.Tech, 4th
year B.Tech, 4th
year
Electronics and Instrumentation Engineering Electronics and Instrumentation Engineering
Galgotias College of Engineering & Technology
Greater Noida
Galgotias College of Engineering & Technology
Greater Noida
cuckoosati@gmail.com devarshi08@gmail.com
Ms Usha Tiwari
Assistant Professor
Dept of Electronics & Instrumentation Engineering
Galgotias College of Engineering & Technology
Greater Noida
ushapant@rediffmail.com
Abstract: In this paper, the design of the
inexpensive high quality panoramic camera has
been discussed. It has various applications in
biomedical, architectural, cartography, scanning
and other fields. With lowered costs, such systems
would have tremendous usage in many more
innovative applications. Though large amount of
research has been done on development of various
algorithms for seamless stitching of images still
some more aspects need to be explored. In this
sceneriao, rigorous work has been done on issues
that affect the quality of the stitched image and
also on the computational complexity of the
algorithms used to address those issues. Instead of
trying to resolve the issues with complicated
computation intensive algorithms, we worked on a
combination of hardware and less computation,
intensive approach to achieve high quality
panoramic pictures. In this paper it has been
demonstrated that image stitching can be
performed using routines and algorithms that
can operate in real time mode to generate
high quality panoramic images. This can be
done by using simple motor controlled camera
that records the parameters required for
image processing algorithms. It can be used
for testing and developing new algorithms.
These can be easily integrated into low cost camera
hardware for generating high quality panoramic
images on the fly.
Keywords: Image Processing, panoramic images,
I. INTRODUCTION
Image stitching is the process of combining
multiple photographic images with overlapping
fields of view to produce a segmented panorama
or high-resolution image. Commonly adopted
approaches to image stitching through the use of
computer software, require nearly exact overlaps
between images and identical exposures to
produce seamless results.
The steps in image stitching generally are as
follows:-
Creation of overlapped images
Motion Models
Image Registration
Blending
Extensive research has gone in each of these
steps. An important step is to establish the
mathematical relationships that can be used to
map pixel coordinates from one image to the
next.
The complexity of the mathematical models
depends on the way the pictures have been taken
and other image characteristics.
2. A. Motion Models
A wide variety of transforms such as simple
2D transforms, 3D transforms and their mapping
to non planar surfaces (e.g. cylindrical, spherical)
are used in the process. The commonly used
transformations are:
Translation
Rigid(Euclidean)
Similarity
Affine
Projective
Each of these translations preserves certain
characteristics of the image such as orientation,
lengths, angles, parallelism and straight lines.
B. Image Registration
One of the oldest and commonly used
algorithm is based on patch based alignment and
developed by Lucas and Kanade [1].
Sophisticated image registration algorithms have
been developed for medical imaging and remote
sensing applications by Goshtashby [2]. For
photogrammetry applications [8] manually
intensive methods based on ground control
points or manually registered tie points have
been used by Slama [3]. The globally consistent
solutions were achieved using the bundle
adjustment algorithm by Triggs et al. [4]. Many
researchers such as Davis [5], Uyttendaele et al.
[6], Agarwala et al. [7] has also presented in
globally consistent alignments and removal of
ghosts due to parallax and object movement.
Another class of algorithms is the feature based
algorithms by Brown et al and Badra et al [10-
11]. Both the above set of algorithms have the
ability to recognize panoramas among an
unordered set of pictures and can be used for
fully automated stitching. Fully automated
stitching algorithms are more computation
intensive and as the number of images increases
the processing requirements increase
significantly.
C. Blending
After identifying the image sequence, amount
of overlap a right blending technique is required
to blend the images seamlessly. Commonly used
blending techniques are laplacian pyramid
blending, gradient domain blending, exposure
compensation, high dynamic range imaging etc.
An alternative approach to motion-based de-
ghosting was proposed by Kang et al. [8], who
estimated dense optical flow between each input
image and a central reference image.
II. PROBLEM ANALYSIS
If various algorithms that are used in the
image stitching domain are looked upon, it is
observed that most of them are attempting to
tackle the issues arising out of images not
conducive to images stitching. Some of the
issues identified are as follows -
Images taken with excessive, low
overlap, undetermined overlap.
The images taken with cameras having,
barrel, pincushion, fisheye distortion.
Exposure differences in pictures taken
i.e. bias and gain.
Translational shift between two images.
Rotational shift between images.
Varying scales and aspect ratios of
images
Focal Length changes
Gaps and Overlaps
Parallax and moving objects
III. OUR APPROACH
Since a lot of research has already been done
in all the aspects of image stitching and proven
algorithms exist for different situations. To get
an accurately stitched image, the algorithms need
a set of pictures , picture parameters and the
desired result. These parameters are used to
identify the correct algorithm and to pass the
requisite parameters for an accurate result to be
used in a particular scenario. The parameters that
affect the choice of the algorithms have been
controlled and recorded. As all pictures are now
taken in a controlled environment with the
parameters recorded, the algorithms now do not
have to estimate the parameters governing their
operation. This implies that we can now select
the right algorithms and pass the right
parameters to them. As a result the images can
be stitched very accurately and the desired
panorama pictures can be produced.
The block diagram of the resulting system is
as shown in figure 1.
3. Figure 1.System Block Diagram
The camera is mounted on a 2 Axis stand
with 2 stepper motors controlling the movement
of the camera in X and Y axis. These steppers
motors are controlled with a computer where
motor speed and the angle of rotation increments
are defined. The camera settings can be adjusted
as desired either with a computer interface or by
directly using the buttons given on the camera .
The choices of parameters are determined by the
features desired in the stitched image. Proper
understanding of the algorithm and the effect on
the resultant image is required to set the
parameters accurately.
The various parameters used in taking the
pictures are recorded as Meta data within the
image or can even be stored as a part of the
filename of the image.
IV. EXPERIMENTAL SETUP
The setup consists of the following:-
Web cam
Camera mount
Regulated Power Supply-9V DC
Laptop
Many digital cameras have interface that can
control the focal length (zoom), aperture,
contrast and color compensation using a
computer interface.
A simple computer Interface was developed
in VB 6. The objective is to control the picture
quality, amount of over lap, eliminate gaps and
record the parameters to be passed on the image
stitching algorithms. This interface uses imaging
, webcam, controls to capture and display the
images. The information thus collected is used to
generate the suitable MATLAB code for
execution. The resultant file after processing in
MATLAB is then re input into the VB6 inteface
for visualization.
A. The camera stand
The camera is mounted on a stand as
depicted in figure 2. This allows for control
and movement of the camera in 2 –Axis.
Precise control and calibration of the
movement mechanism is done using stepper
motors and the vital parameters required for
stitching of images can be controlled and
recorded.
Figure 2. Camera Stand
B. The interface
The interface as designed in figure 3 allows
the user to control the image capture process and
to display the various pictures as they are taken
and also the resultant stitched image. The
pictures taken of our college (Galgotias College
of Engineering and Technology) were stitched
together to demonstrate the results.
4. Figure 3.Interface
C. Operation of the Camera
The camera settings are adjusted as per
requirement by clicking on the camera settings
button. Once the user starts the camera by
clicking on start camera the current image being
clicked is shown as in figure 4.
Figure 4.Camera Controller
The camera movement parameters can be set
as shown in the interface in figure 5 for each
axis. Consecutive pictures are taken and saved
into a folder as specified in the computer
interface.
Figure 5
The pictures thus taken are then stitched
using appropriate MATLAB functions.
MATLAB has an image processing library
which we have used to stitch the images. The
functions are used to calculate the Transform
Matrix, the RANSAC algorithm verifies that the
corners have been identified correctly and the
Mosaicking subsystem to overlay the frames
correctly.
Figure 6. Image after Stitching
V. RESULTS
Various pictures of moving and non moving
objects were taken and stitched together using
various algorithms and the resultant images were
analyzed. It was observed that with proper
settings of the camera, high quality results were
possible even with the simplest of the stitching
algorithms. When the seams of the images had
moving objects ghost images were observed. The
ghost images could be reduced by decreasing the
overlap when the pictures had moving objects
near the seams without using ghost removal
algorithms.
This essentially paves the way for developing
an inexpensive portable camera system that can
take panoramic images of high quality using
ordinary camera technology and simple image
stitching algorithms that can be integrated into a
low end microprocessor.
5. Sophisticated images of high quality can be
generated using multiple cameras controlled
centrally or over a network as in case of
recording of games. The pictures thus acquired
can be processed by more sophisticated
algorithms with cameras spread over various
physical positions.
VI. CONCLUSION
In this paper it has been demonstrated that
image stitching can be performed using routines
and algorithms that can operate in real time
mode to generate high quality panoramic images.
This can be done by using simple motor
controlled camera that records the parameters
required for image processing algorithms. It can
be used for testing and developing new
algorithms. In this paper various options have
been considered to work out a cost effective and
modular image stitching camera. The objective is
to establish a strong base for further refinement
and improvement in the image processing and
multidisciplinary applications of the technology.
Advances and refinements in camera technology
coupled with improvements in image processing
technology can easily be integrated into the
highly modular design of the system.
Large format projection, printing and
scanning applications will find great use of this
technology. Medical imaging costs can be
reduced significantly by extending concepts used
in this paper.
VII. REFERENCES
[1] Lucas, B. D. and Kanade, T.
(1981). An iterative image registration
technique with an application in stereo
vision. In Seventh International Joint
Conference on Artificial Intelligence
(IJCAI-81), pages 674–679, Vancouver.
[2] Goshtasby, A. (1989). Correction
of image deformation from lens
distortion using Bezier patches.
Computer Vision, Graphics, and Image
Processing, 47(4), 385–394.
[3] Slama, C. C., editor. (1980).
Manual of Photogrammetry. American
Society of Photogrammetry Falls Church,
Virginia, fourth edition.
[4] Triggs, B. et al.. (1999). Bundle
adjustment — a modern synthesis. In
International Workshop on Vision
Algorithms, pages 298–372, Springer,
Kerkyra, Greece.
[5] Davis, J. (1998). Mosaics of scenes
with moving objects. In IEEE Computer
Society Conference on Computer Vision
and Pattern Recognition (CVPR’98),
pages 354–360, Santa Barbara.73
[6] Uyttendaele, M., Eden, A., and
Szeliski, R. (2001). Eliminating ghosting
and exposure artifacts in image mosaics.
In IEEE Computer Society Conference
on Computer Vision and Pattern
Recognition (CVPR’2001), pages 509–
516, Kauai, Hawaii.
[7] Agarwala, A. et al.. (2005).
Panoramic video textures. ACM
Transactions on Graphics, 24(3), 821–
827.
[8] Frizot, Michael (ed.), 1998: Neue
Geschichte der Fotografie. Könemann
Verlagsgesellschaft, Köln.von Gruber, O.
(ed.), 1930: Ferienkurs in
Photogrammetry. Verlag Konrad
Wittwer, Stuttgart, 510p.
[9] Kang, S. B. et al.. (2003). High
dynamic range video. ACM Transactions
on Graphics, 22(3) 319–325. Seon Joo
Kim, Marc Pollefeys, "Robust
Radiometric Calibration and Vignetting
Correction," IEEE Transactions on
Pattern Analysis and Machine
Intelligence, Apr. 2008 vol. 30(4), pp.
562-576,
[10] Brown, M., Szeliski, R.,
andWinder, S. (2005). Multi-image
matching using multi-scale oriented
patches. In IEEE Computer Society
Conference on Computer Vision and
Pattern Recognition(CVPR’2005), pages
510–517, San Diego, CA.
[11] Badra, F., Qumsieh, A., and Dudek,
G. (1998). Rotation and zooming in
image mosaicing. In IEEE Workshop on
Applications of Computer Vision
(WACV’98), pages 50–55, IEEE
Computer Society, Princeton