This paper discusses techniques for digital image processing, including noise reduction, edge detection, and histogram equalization. Noise reduction techniques discussed include mean, Gaussian, and median filters to remove salt and pepper noise and Gaussian noise. Edge detection algorithms like Sobel and Laplacian are introduced to reduce image data while preserving object boundaries. Histogram equalization is used for image enhancement by spreading pixel values across the full intensity range for increased contrast. The goal is recognizing objects in images through these preprocessing steps.

1. Recognization of Objects
In
Real Time
Saloni Bhatia, Ambuj Gupta
Electronics andcommunication department
Guru Jambeshwar University of Science and Technology
Hisar, Haryana
salonibhatia13@gmail.com
ambkish@gmail.com
Abstract— This paper is a result of rigorous work of 2 months and
deals with the various methods involved in image processing. Given
an image of considerable size and applying major techniques and
algorithms including edge detection, histogram, noise reduction to
make an image appealing. Thresholding technique used results in
the recognization of the desired object.
Keywords— image processing, thresholding, edge detection,
noise reduction
I. INTRODUCTION
Image processing involves altering an existing image in a
desired manner. There are basically two methods of
processing images that is optical processing and electronic
processing with later being further categorized as analog and
digital processing. This paper deals with digital image
processing techniques which uses picture elements or pixels.
The required property of image is extracted from processing
an image either in black and white or in a colour and in a print
form or in digital form.
Firstly problem is considered and solution is obtained by
passing through following stages:
• Image acquisition dealing with capturing of image by a
sensor and digitized, by using analog to digital
converter.
• Recognization of object by labelling it based on
information provided and then interpreting to ensemble
recognized objects.
• Segmentation of information contained in the image
into smaller entities.
• Representation and Description transforms raw data
into a form suitable for recognization.
• Knowledge base where the information of interest is
known to be located to limit the search.
In section II we deal with the various techniques and
algorithms involved in removal of noise from the image.
In section III the algorithms involving edge detection are
introduced to reduce the amount of data. The next section,
section IV, deal with histogram matching and its
equalisations. Section V deals with the process of image
segmentation including thresholding.
II. NOISE REDUCTION
Noise is a major factor that influence our image produced
during process of image acquirement and transmission. Image
taken with both digital and conventional film cameras will
pick up noise from variety of sources. Since working of image
transmitter is influenced by various factors hence various
noise reduction techniques are present to deal with different
noises.
There are basically two types of noises dealt in this paper.
The first kind is salt and pepper noise in which some pixels in
image are very different in colour or intensity from
surrounding pixels resulting in a dark and white dots. The
second one is Gaussian noise in which each pixel in an image
will be changed from its original value by a small amount.
Various methods are applied to reduce these disturbances.
A. Mean Filter
It is a linear, low-pass filter where all coefficients have
identical values. It’s characterstics depends upon the kernel
width, height and shape. If size of the kernel increases the
smoothing effect also increases. Selection of kernel size and
form is a compromised between reduction of noise and a low
blurring effect. A mean 3*3 rectangular filter is defined by:
Where 1/k is scaling factor with k=mean=9
Fig. 1 Example of mean filter with test image and result image
B. Gaussian Filter

2. It is a linear filter used to remove Gaussian noise by
convolving the original image with a mask that represents a
Gaussian function. Convolution brings the value of each pixel
closer with the values of its neighbours. A matrix 3*3
rectangular Gaussian filter is given by
; k = 9
Where k= sum of the elements of matrix
Fig. 2 Example of Gaussian filter with test image and result image
Note: smoothing filters tends to blur an image because
pixel intensity values that are higher or lower than
surrounding neighbourhood would smear across the area.
Hence, non-linear filters are mostly used.
C. Median Filter
It is a non-linear filter which is an excellent choice for the
removal of shot noise and horizontal scanning artefacts.
To perform this filter a data window is move over an entire
image. If a centre value of data window differs more than the
certain value it will be exchanged by median value. The
threshold value depends upon the local standard deviation
within data window. If standard deviation is increased the
threshold value also increases.
Fig.3 Example of median filter with test image and result image
III. EDGE DETECTION
This operation on images drastically reduces the amount of
data to be processed and preserves structural information
about object boundaries. It basically uses laplacian and
gradient methods. While Gradient method uses first derivative
to detect edges, the Laplacian uses second derivative to search
zero crossing. The egde detection operator basically
determines the variations in pixel values. A threshold value is
set and if the pixel value in an image is above it than that area
is classified as an edge. The major methods include
A. Sobel edge detection
It is used to perform a 2-D gradient measurement. It
determines the gradient for columns and rows separately using
3*3 convolution masks which is given by
Thus gradient
The mask is made to move over the entire area of a image
and changes the pixel’s value and then shifts one pixel to right
and continues till the entire pixels are manipulated.
Fig.4 Example of Sobel edge detection with test and result image
B. Laplacian edge detection
This operator uses only one mask for both horizontal and
vertical directions. It is used as an approximation for second
order derivative, due to which this mask becomes very
sensitive to noise and also produces double edges. Thus it is
used to determine whether pixel value lies within dark or light
side of an edge.
Laplacian operator is not used in its original form.
Fig.5 Example of Laplacian edge detection with test and result images

3. IV. HISTOGRAM EQUILIZATION
Histogram shows the total tonal distribution in the image. It
is a kind of a barchart of the count of pixels of every tone of
gray that occurs in the image. It is basically used for image
enhancement and various image processing applications of
image compression and segmentation. For the histogram of an
image equal sized classes are taken where vertical axis
represents frequency and horizontal axis represents no. of
pixels. It is the colour adjustment of two images whereas
histogram equilization is a method of contrast adjustment
using the image’s histogram.
Fig.6 Example of histogram equalization with test and result image
In above given images components of the Histogram of dark
image are concentrated on lower side of intensity scale.
Similarly components of histogram of light image are
concentrated toward the high intensity side. An image with
low contrast has a narrow histogram located toward middle of
the scale. Finally, in high contrast image components of
histogram cover a wide range of intensity scale. Thus the
image whose pixels occupy the entire range of possible
intensity level tend to distribute uniformly have an appearance
of high contrast images and exhibit a large variety of gray
levels. Thus the histogram equalization provides us with high
contrast images since it spread pixels over the entire range of
intensity.
ACKNOWLEDGMENT
The heading of the Acknowledgment section and the
References section must not be numbered.
Causal Productions wishes to acknowledge Michael Shell
and other contributors for developing and maintaining the
IEEE LaTeX style files which have been used in the
preparation of this template. To see the list of contributors,
please refer to the top of file IEEETran.cls in the IEEE LaTeX
distribution.
REFERENCES
[1] S. M. Metev and V. P. Veiko, Laser Assisted Microtechnology, 2nd
ed., R. M. Osgood, Jr., Ed. Berlin, Germany: Springer-Verlag, 1998.
[2] J. Breckling, Ed., The Analysis of Directional Time Series:
Applications to Wind Speed and Direction, ser. Lecture Notes in
Statistics. Berlin, Germany: Springer, 1989, vol. 61.
[3] S. Zhang, C. Zhu, J. K. O. Sin, and P. K. T. Mok, “A novel ultrathin
elevated channel low-temperature poly-Si TFT,” IEEE Electron
Device Lett., vol. 20, pp. 569–571, Nov. 1999.
[4] M. Wegmuller, J. P. von der Weid, P. Oberson, and N. Gisin, “High
resolution fiber distributed measurements with coherent OFDR,” in
Proc. ECOC’00, 2000, paper 11.3.4, p. 109.
[5] R. E. Sorace, V. S. Reinhardt, and S. A. Vaughn, “High-speed digital-
to-RF converter,” U.S. Patent 5 668 842, Sept. 16, 1997.
[6] (2002) The IEEE website. [Online]. Available: http://www.ieee.org/
[7] M. Shell. (2002) IEEEtran homepage on CTAN. [Online]. Available:
http://www.ctan.org/tex-
archive/macros/latex/contrib/supported/IEEEtran/
[8] FLEXChip Signal Processor (MC68175/D), Motorola, 1996.
[9] “PDCA12-70 data sheet,” Opto Speed SA, Mezzovico, Switzerland.
[10] A. Karnik, “Performance of TCP congestion control with rate
feedback: TCP/ABR and rate adaptive TCP/IP,” M. Eng. thesis, Indian
Institute of Science, Bangalore, India, Jan. 1999.
[11] J. Padhye, V. Firoiu, and D. Towsley, “A stochastic model of TCP
Reno congestion avoidance and control,” Univ. of Massachusetts,
Amherst, MA, CMPSCI Tech. Rep. 99-02, 1999.
[12] Wireless LAN Medium Access Control (MAC) and Physical Layer
(PHY) Specification, IEEE Std. 802.11, 1997.