Professional Documents
Culture Documents
120
ITSI Transactions on Electrical and Electronics Engineering (ITSI-TEEE)
small amount of noise is harmful when high accuracy is very similar regardless of the signal being processed
required, e.g. as in sub cell (subpixel) image analysisIn [14], however a priori knowledge of the characteristics
this paper, we propose to denoise images by filtering the of an expected signal can mean the implementations of
image. The Box filter is used for removing the noise these techniques vary greatly depending on the type of
from the image. signal. In practice a lot of methods are used to eliminate
the noise from the image and a lot of filters are used. In
II. RELATED WORKS
this paper we used box filter for removing the noise
Literature survey in [1][2][4] addressed various from the image
issues regarding the challenges faced by research
III. IMAGE TRANSMISSION USING BOX FILTER
community in realizing WMSN. The early research
IN WMSN
efforts in wireless sensor networks did not investigate
the issues of node collaboration, focusing more on 3.1 Lossless Image Compression using Improved
issues in the design and packaging of small, wireless Polyomino
devices [5], more recent efforts (e.g. [6], [7]) have
considered node collaboration issues such as data There are two types of image compression: lossless
“aggregation” or “fusion”. Our approach of distributed and lossy. After decompression the original image is
image compression falls within the domain of recovered. Compressing an image is significantly
techniques that apply the concept of in-network different than compressing raw binary data. The general
processing, i.e. processing in the network by computing purpose compression is used to compress images, but
over the data as it flows through the nodes. It is worth the result is less than optimal. This is because images
noting that current aggregation functions (e.g., have certain statistical properties which can be exploited
“maximum” and “average” [7]) are limited to scalar by encoders specifically designed for them. This also
data. Our approach can be viewed as an extension to means that lossy compression techniques can be used in
vector data aggregation. this area. An integer-to-integer wavelet transform
produces an integer-valued transform from the grey-
Previous distributed signal processing/compression scale, integer-valued image. Since n loops in Bit-plane
problems (e.g. [8], [9]) exploit correlations between data encoding reduces the quantization error to less than
at close-by sensors in order to jointly compress or fuse T0/2n, it follows that once 2n is greater than T0, and
the correlated information resulting in savings in there will be zero error. In other words, the bit-plane
communication energy. In parallel distributed encoded transform will be exactly the same as the
computing theory [10], a problem (or task) is divided original wavelet transform, hence lossless encoding is
into multiple sub-problems (or sub-tasks) of smaller size achieved Lossless compression involves with
(in terms of resource requirements). Every node solves compressing data which, when decompressed, will be an
each sub problem by running the same local algorithm, exact replica of the original data. This is the case when
and the solution to the original problem is obtained by binary data such as executables, documents etc. are
combining the outputs from the different nodes. Our compressed. They need to be exactly reproduced when
approach to the design of distributed image compression decompressed.
is similar in concept, in that we distribute the task of
image encoding/compression to multiple smaller image 3.2 Wavelet Image Compression
encoding/compression sub-tasks. our proposed approach
of image compression intersects with the literature on Wavelet based image compression introduces no blocky
lossy and lossless compression, which primarily focuses artifacts in the decompressed image. It is decompressed
on polyomino technique. image is much smoother and pleasant to eyes. We can
also achieve much higher compression ratios regardless
Digital images are prone to a variety of types of
of the amount of compression achieved. By adding more
noise. There are several ways that noise can be
introduced into an image, depending on how the image and more detail information we can improve the quality.
is created. For example: If the image is scanned from a This feature is attractive for what is known as
progressive transmission of images. Another lossy
photograph made on film, the film grain is a source of
compression scheme developed for image compression
noise [11]. Noise can also be the result of damage to the
is the fractal base image compression scheme (fig 1).
film, or be introduced by the scanner itself. If the image
However the fractal based image compression beginning
is acquired directly in a digital format, the mechanism
for gathering the data (such as a CCD detector) can to loss ground because it is very complex and time
introduce noise. Electronic transmission of image data consuming.
can introduce noise [12]. Image noise elimination
The filter components are reduced their size by half
(reduction) [13] is the process of removing noise from
either by rejecting the even or odd samples thereby the
the image. Noise reduction techniques are conceptually
total size of the original signal is preserved. The low sufficient energy to forward them, if it knows that a
pass filter component retains almost all distinguishable node further down the path has an insufficient report the
features of the original signal. And the high pass filter lowest energy level currently available in others nodes.
component has little or no resemblance of the original The delay induced by the feedback is proportional to the
signal. The low pass component is again decomposed distance between the concerned nodes .
into two components. The decomposition process can be
continued up to the last possible level or up to a certain 3.4 Noise Reduction Filter
desired level. As the high pass filter components have
less information discernible to the original signal, we The filters in the Noise Reduction class are
can eliminate the information contents of the high pass designed to remove extreme or outlier values from the
filters partially or significantly at each level of transmitted image areas that should have relatively
decomposition during the reconstruction process. It is uniform values. These outlier values are often the result
this possibility of elimination of the information of additive noise imposed on the image by the
contents of the high pass filter components that gives acquisition system or later processing errors. Median,
higher compression ratio in the case of wavelet based Modal etc., are some of the examples of this type. In this
image compression. work a study is made on the box filtering techniques
used to remove the noise from the transmitted image by
3.3 Energy Efficient High Quality Image Transmission EEHQIT Scheme.
scheme
A spatial averaging filter in which all coefficients
In this paper, we propose image transmission are equal is called a box filter. These types of filters are
scheme driven by energy efficiency considerations in used for blurring and for noise reduction. The output of
order to be suitable for wireless sensor networks. such a linear smoothing filter is simply the average of
Wavelet image transform provides data decomposition the pixels in the neighborhood of the pixel mask. The
in multiple levels of resolution, so the image can be idea behind smoothing filters is straight forward. By
divided into packets with different priorities, the packets replacing the value of every pixel in an image by the
are ready to be sent. average of the levels in the neighborhood of the filter
mask, the process results in an image with reduced
„sharp transitions‟. Hence the most obvious application
of smoothing is noise reduction.
computed the average energy consumption to transmit above approximately 50 dB, although this may vary in a
the image for scenario (Interactive Image Transmission minor way for each person
Energy and Time WT).The parameters which are used
in the filter performance evaluation are Signal to Noise
Ratio (SNR), Peak Signal to Noise Ratio (PSNR).
Calculation of SNR
Vo2
SNR 10 log
Ve2 Figure 2: Imge size vs PSNR
V02 is the variance of the noise free image and Figure 2 depicts the performance graph for showing the
Where, result of PSNR value at sender side and receiver side.
Ve2 is the variance of error (between the original and PSNR value is measured in terms of decibel. At the
receiving side, the recovered DC and AC components
denoised image). Brighter regions have a stronger signal are used to reconstruct the image. When channel noise is
due to more light, resulting in higher overall SNR 38 not considered, the image quality keeps improving, in
Calculation of PSNR terms of its PSNR value, as we use more AC
PSNR is the ratio between possible power of a signal components. However, to achieve some degree of
and the power of corrupting noise that affects the compression, some of the AC components, those that
fidelity of its representation. correspond to the high frequencies in the image, can
PSNR in decibels is easily defined from RMSE as given safely be excluded when reconstructing the image. By
below using our proposed technique the PSNR value of
PSNR = 20 log10 (255/RMSE) receiver side is nearly equal to the sender side value.
Where,
1 M N
RMSE [ I ( x, y) I ' ( x, y)]2
MN y 1 x 1
where I(x,y) is the original image, I'(x,y) is the
decompressed image and M,N represents dimensions of
the images. A lower value for MSE means lesser error,
and as seen from the inverse relation between the MSE
and PSNR, this translates to a high value of PSNR.
Logically, a higher value of PSNR is good because it
means that the ratio of Signal to Noise is higher. The
signal is the original image, and the noise is the error in
reconstruction. It is highly required to evaluate a
compression scheme having a lower MSE (and a high
PSNR). Figure 3: No. Of Images vs PSNR
V. RESULT AND DISCUSSIONS Figure 3 gives the performance graph for showing the
As a performance measurement, the Peak Signal-to- result of PSNR value at sender side and receiver side
Noise Ratio (PSNR) is calculated for the reconstructed with number of images used. By using our proposed
images at the receiver side. PSNR is metric used to technique the PSNR value of receiver side is nearly
compare two images, the more pixel difference between equal to the sender side value.
the images, the less the PSNR value. It is useful to know
that the human eye does not have enough sensitivity to
detect changes in visual data for PSNR measurements
[3] W. Zhang, Z. Deng, G. Wang, L. Wittenburg, and [12] Ricardo G. Dantas and Eduardo T. Costa,
Z. Xing, “Distributed problem solving in sensor “Ultrasound Speckle Reduction Using Modified
networks,” in Proceedings of the first Gabor Filters”, IEEE TRANSACTIONS ON
international joint conference on Autonomous ULTRASONICS, FERROELECTRICS AND
agents and multiagent systems. ACM Press, FREQUENCY CONTROL, 2007, VOL.54,
2002, pp. 988–989. pages: 530-538
[4] Liu, C.-M., Lee, C.-H., and Wang, L.-C.. Power- [13] F. Labeau, J. C. Chiang, M. Kieffer, P. Duhamel,
e±cient communication algorithms for wireless L. Vandendorpe, and B. Marq. Oversampled filter
mobile sensor networks. In 1st ACM banks as error correction codes: Theory and
International Workshop on Performance impulse noise correction. submitted to IEEE
Evaluation of Wireless, Ad Hoc, Sensor and Trans. on Signal Processing, 2004.
Ubiquitous Networks, 2004 pages 121-122. [14] G. Rath and C. Guillemot. Characterization of a
[5] Magli, Mancin, M., and Merello, L.. Low class of error-correcting frames and their
complex- ity video compression for wireless application to image transmission. In Proceedings
sensor networks. In Proceedings of 2003 of PCS, St Malo, 2003
International Conference on Multi- media and
Expo, 2003 pages 585-588.
[6] W. Yu, Z. Sahinoglu, and A. Vetro, “Energy
efficient JPEG 2000 image transmission over
wireless sensor networks,” in Proceedings of
IEEE Global Telecommunications Conference