Gv2512441247

Swamy.TN, Rashmi. KM, Dr.P.Cyril Prasanna Raj, Dr.S.L.Pinjare / International Journal of
Engineering Research and Applications (IJERA) ISSN: 2248-9622 www.ijera.com
Vol. 2, Issue 5, September- October 2012, pp.1244-1247
FPGA Implementation Of Efficient Algorithm Of Image Splitting
For Video Streaming Data
Swamy.TN*, Rashmi. KM**, Dr.P.Cyril Prasanna Raj ***, Dr.S.L.Pinjare
****
*(Assistant Prof, Rajiv Gandhi Institute of Technology, Bangalore)
** (Assistant Prof, SDMIT, Ujire, Mangalore)
***(Director, NXG semiconductor Technologies, Bangalore)
****(PG coordinator, Nitte Meenakshi Institute of Technology, Bangalore)

ABSTRACT
Video splitting is the process of dividing it can be processed as a steady and continuous
the video into non overlapping parts. Then row stream over the network. With streaming the client
mean and column mean or each part is obtained. browser or plug in can start displaying the
After applying transform on these, features sets multimedia data before the entire file has been
can be obtained to be used in image retrieval. By transmitted.
using splitting higher precision and recall can be
obtained. Streaming refers to transferring video
data such that it can be processed as a steady and
continuous stream over the network. With
streaming the client browser or plug in can start
displaying the multimedia data before the entire
file has been transmitted. In this paper original
image of particular size is split into blocks. Each
block is enlarged to the original dimension
without blurring. Enlarged image is displayed on
the big screen. This can be done on video
streaming data. Designs are implemented using
Xilinx and Mat lab tools and Xilinx Vertex-2p
FPGA board.

Keywords– Croma, interpolation, luma,
simulink, system generator.

I. INTRODUCTION Fig1. Original image and splitted image
1.1 Video splitting
Video splitting is the process of dividing 1.2 Interpolation
the video into non overlapping parts. Then row The goal of interpolation is to produce the
mean and column mean of each part is obtained. acceptable images at different resolution from a
After applying transform on these, feature sets can single low resolution image. The actual resolution of
be obtained to be used in image retrieval. By using the image is defined as the number of pixels. Image
splitting higher precision and recall can be obtained. interpolation is a basic tool used extensively in
Natural images are captured using image sensors in zooming, shrinking, rotating and geometric
the form of voltage intensities that are digitized and corrections. Interpolation is the process of using
stored in memory banks. Large storage space is known data to estimate values of unknown
required to store and process these digital samples locations, suppose that an image of size 500*500
[1]. For example, a color image of size 256*256 pixels has to be enlarged 1.5 times to 750*750
represented using 24-bit requires a storage space of pixels. A simple way to visualize zooming is to
47 Mega bits. Similarly the storage space for a three create an imaginary 750*750 grid with the same
hour movie requires 92000 Giga bits. pixel spacing as the original, and then shrink it so
Processing and transmission of huge image that if fits exactly over the original image.
data is time consuming and very cumbersome. Obviously, the pixel spacing in the shrunken
Hence the project has a major application in all 750*750 grid will be less than the pixel spacing in
areas of image processing. Fig1 shows original and the original image [2]. To perform intensity-level
spitted non overlapping parts respectively. assignment for any point in the overlay, we look for
Streaming refers to transferring video data such that its closest pixel in the original image and assign the

1244 | P a g e

intensity of that pixel to the new pixel in the
750*750 grid. When we are finished assigning
intensities to all the points in the overlay grid, we
expand it to the original size to obtain the zoomed
image. This method is called nearest neighbor
interpolation because it assigns to each new location
the intensity of its nearest neighbor. The basic
criteria for a good interpolation method are
geometric invariance, contrast invariance, noise,
edge preservation, aliasing, texture preservation,
over smoothing, application awareness, and
sensitivity to parameters [3].
Fig3. Image matrix to serial format
II. DESIGN
The block diagram of the design is shown The conversion of serial data into matrix
in Fig2. Video from a web camera at 30 frames per format is shown in Fig4. The hardware design using
second is applied to the video input block. Resize the system generator is shown in Fig5.
block enlarge or shrinks image size. Captured video
will be in RGB format. Input video is converted into
croma and luma components.
Luma represents the brightness in an image
and it represents the achromatic image without any
color while the croma component represents the
color information. Image split block splits the image
into number of blocks. Each spitted block is resized
using bicubic interpolation technique. The split
image is displayed using the video display output
block.

Fig4. Image serial to parallel conversion
Video Resize RGB to
input 256*256 YCrCb

Splitting

Video Resize YCbCr to
output 256*256 RGB

Fig5. System generator model
Fig2. Block diagram of image splitting
III. PARAMETERS
The design is implemented using simulink and 3.1 Video input block
system generator [3][4] environment. 3.1.1 Device
Image will be in matrix format. It has to be Device describes the image acquisition
converted into serial data before applying into the device to which we want to connect. The items in
system generator block. The design for converting the drop down list of device vary depending on
matrix to serial format is as shown in Fig3. which devices we have connected to the system. All
Transpose block is used to convert M*N video capture devices supported by image
matrix into N*M matrix. The block named convert acquisition toolbox are supported by the block.
2D to 1D block reshapes an M*N matrix into a 1D
vector with length M*N. Frame conversion block 3.1.2 Video format.
specifies the frame status of the output signal. The Video format shows the video formats
unbuffer block unbuffers frame based input into supported by the selected device. The drop down list
sample based output. of video format varies with each device. If the
device supports the use of camera files, from camera
file will be one of the choices in the list.

1245 | P a g e

The input image/video is splitted into four blocks of
3.1.3 Block sample time. dimension 128*128 each as shown in Fig7.
It specify the sample time of the block
during the simulation. This is the rate at which the
block is executed during simulation. The default is
1/30.

3.1.4 Ports model.
It is used to specify either a single output
port for all color spaces, or one part for each band
(for example R, G, B). When you select one
multidimensional signal, the output signal will be
combined into one line consisting of signal
information for all color signals. Select separated
color signals if you want to use three ports Fig7. Splitted image(128*128*3)
corresponding to the uncompressed red, green, and
blue color bands. These splitted images are interpolated to the original
dimension of the input, i.e. 256*256. Bicubic
3.1.5 Data type interpolation is used to resize the splitted image into
Image data type when the block outputs the original dimension. The resized image is of good
frames. This data type indicated how image frames quality because it resembles almost the resolution of
are output from the block to simulink. It supports all the original image. The interpolated images are
Matlab data types and single is the default. shown in Fig8.

3.2 Color space conversion
The color space conversion block converts
color information between color spaces. Conversion
parameter is used to specify the color spaces we are
converting. Conversion between RGB and YCbCr
and vice-versa are defined by the following
Equation (1).

(1)
Fig8. Interpolated image(256*256*3)
IV. RESULTS
4.1 Simulink output 4.2 Hardware output
The input image is in RGB format and of size The RGB image of dimension 128*128*3
256*256*3 shown in Fig6. is considered as input image as shown in Fig9.

Fig9. Input image for hardware(128*128*3)

Fig6. Original image 256*256*3 The image is resized to 128*128 and converted into
gray image as shown in Fig10.

1246 | P a g e

DSP with xilinx system generator for
Matlab, processdings of 35th south eastern
symposium, Vol 15, page 2226-2238, 2006
[1] [5]Akhtar. P and Azhar. F, A single image
interpolation scheme for enchanced super
resolution in Bio-Medical Imaging, 4th
International Conference on Bio-
informatics and Bio medical Engineering,
pages 1-5, June 2010.
Fig10. Gray image(128*128) [6] Fatma T. Arslan and Artyom M.
Grigoryan, Method of image enhancement
The output on the FPGA hardware is as shown in by splitting signal, IEEE international
Fig11. Conference on Acoustics, speech and
signal processing, vol 4, page iv/177 -
iv/180, March 2005
[7] Rafael C. Gonzalez and Richard E. Woods,
Digital Image Processing, Pearson Prentice
Hall, 3 Edition, 2008.
[8] Samir Palnitkar, Verilog HDL a Guide to
Digital Design and Synthesis, Pearson
Education, 2 Edition, 2009.

Fig11. Hardware output image(128*128)

V. CONCLUSION
This paper is very useful in displaying
wider images for commercial purposes in
Aerodromes and railway stations. And can be
widely used in medical fields. In medical field
doctor can easily indentify the affected area by
zooming that part without zooming entire image.

REFERENCES
[1] Ian kuon and Jonathan Rose, Measuring the
gap between FPGAs and ASICS, IEEE
transcation on Computer-Aided of
Integrated Circuits and systems, 26(2):
203-215,2007.
[2] Marco Aurelio,Nunu maganda, Real time
FPGA based architecture for bicubic
interpolation: an application for digital
image scaling. Proc of the International
conference on reconfigurable computing
and FPGAs, 2005
[3] T.Saidani, D. Dia, W. Elhamzi, M. Atri,
and R. Tourki, Hardware Cosimulation for
Video Processing using Xilinx System
Generator, Proc of the World Congress on
Engineering, Vol 1, page 1-3, 2009
[4] Matthew Own by, Dr Wagdy.H. mhmoud,
A design methodology for implementing

1247 | P a g e

Gv2512441247

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to Gv2512441247 (20)

Gv2512441247