JOURNAL OF LATEX CLASS FILES, VOL. 14, NO.

8, AUGUST 2023 1

A Comparative Study On Image Compression in

Cloud Computing
Tri Yulianto Nugroho¹, Ahmad Nur Hidayat¹, Mas Syahdan Filsafan¹, Yoehan Arif Ardiansyah¹, and Bagus Jati
Santoso¹
Department of Informatics, Institut Teknologi Sepuluh Nopember, Surabaya, Indonesia
email: [email protected], [email protected], [email protected],
[email protected], [email protected]

Abstract—The purpose of this study is to compare the use of advantages of DWT is its ability to reduce data redundancy
Discrete Cosine Transform (DCT), Discrete Wavelet Transform and maintain picture or sound quality with minimal loss
(DWT), Singular Value Decomposition (SVD), and Karhumen- of information. DWT can also produce smaller file sizes
Loeve Transform (KLT) as image compression techniques in
the Cloud computing. of the four methods we will compare compared to other compression techniques, such as JPEG
the amount of compression, the speed during the compression or MP3. DWT can also be used with lossless compression
process, and compare the image quality after compression. In techniques.[3]
this study there are 6 examples of photos that will be used in When it comes to using cloud computing, speeding up data
testing. transfers is key to ensuring that users get optimal performance.
Index Terms—cloud computing, mage compression One way to speed up data transfer is through lossless com-
pression. Lossless compression is a compression technique in
I. I NTRODUCTION which data can be compressed into a smaller size without
losing data quality.
With the increasing dependence of humans on technology Researchers will also try to use data compression using
from day to day, the resulting digital information is also in- Discrete Cosine Transform (DCT) which is one of the most
creasing and can be accessed anywhere with any device. That commonly used compression techniques. DCT quickly and
reliance can take many forms, from taking photos of beautiful effectively generates frequency coefficients that efficiently
landscapes to collecting images for training in driverless car represent the original signal. In DCT, the original signal is
systems and going beyond creating industrial and medical partitioned into small blocks and each block is converted
images. This generated information whether acoustics, images, into a transformative signal. Each block is then represented
etc. needs to be stored in a media file before being processed by a number of frequency coefficients sorted by frequency
or transmitted. These media files occupy space on the storage magnitude.
device which can be accessed from anywhere. As the number In general, DCT is usually used in data compression appli-
of files increases, we are forced to delete less important files cations to reduce the size of data files. Data compression using
or buy more storage devices[1]. the DCT algorithm is usually lossy because some unimportant
Initially, emphasis of research in image compression was information in the data can be lost. Therefore, the quality of
on the development of analog methods for reducing video data before and after compression can be different.
transmission bandwidth, however, with the advent of digital Researchers will also try several other data compression
computers and subsequent development of advanced integrated methods such as Karhunen-Loéve Transform (KLT) and Sin-
circuits, attention shifted from analog compression to digital gular Value Decomposition (SVD). Where these methods have
approaches. With the development of practical application different image compression features and have their respective
based on several key international image compression stan- advantages and disadvantages. KLT is the most efficient linear
dards, the field has experienced substantial growth. The main transformation for energy compaction. The transformation
purpose of image compression is to store or transmit images matrix can be obtained by calculating the eigenvectors of
in an efficient form. Compression is achieved by lessening the covariance of the input image data to reduce the high
redundancies such as inter-pixel redundancy, psycho-visual computational complexity. While SVD is a linear algebraic
redundancy, and coding redundancy of the image data.[2] technique that is used to break the image into 2, namely the
We will try to use the Discrete Wavelet Transform (DWT) orthogonal matrix and the diagonal matrix. Where SVD is
for image compression. Discrete Wavelet Transform (DWT) is more suitable is used to narrow down the low rating estimates
a data compression technique based on signal processing. This on images.
technique divides the signal into several levels and each level In this study there are 6 examples of photos that will
will be compressed separately. DWT can be used to compress be used in testing. The sample photos will be compressed
various types of data, such as images and sound. One of the using the Discrete Cosine Transform (DCT), Discrete Wavelet
J. Doe and J. Doe are with Anonymous University. Transform (DWT), Singular Value Decomposition (SVD), and
Manuscript received April 19, 2023; revised August 26, 2023. Karhumen-Loeve Transform (KLT) as image compression
JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2023 2

techniques in the Cloud. testing will focus on comparing the which are transformations of vector node to another vector
amount of compression, speed during the compression process, space or a different dimension.[6] SVD is a powerful and
and comparison of image quality after compression. useful tool for decomposition in linear algebra which functions
for image processing such as image encoding and image noise
II. E ASE OF U SE removal. For example, metric A with m rows and n columns,
Related Work will produce orthogonal metrics Y (m x m) and V (n x n), so:

III. METHODOLOGY U = u1 u2 · · · um and V = v1 v2 · · · vn (1)

In this study there are 6 examples of photos that will be can be proven by :
used in testing. we will compare the amount of compression,
T
the speed during the compression process, and compare the U AV = diag (σ1 ,..., σp ) , p = min(m,n) (2)
image quality after compression. There are 4 methods that
where :
will be used to make comparisons: Discrete Cosine Trans-
σ1 ≥ σ2 ,..., σp ≥ 0 (3)
form (DCT), Discrete Wavelet Transform (DWT), Singular
Value Decomposition (SVD), and Karhumen-Loeve Transform ai is the singular value of A and the vector ui and vi is the i-
(KLT) as image compression techniques in Cloud Computing. th left and i-right singular vectors single vector each.[7] SVD
contains intensity image information, modifying a single value
A. Discrete Cosine Transform produces better image contrast.[8]
One of the most often used methods of data compression
is the discrete cosine transform (DCT). Frequency coefficients D. Karhumen-Loeve Transform
produced by DCT efficiently and quickly represent the original
The Karhunen-Loéve transform(KLT) is linear orthogo-
signals. The original signal is divided into smaller blocks for
nal transformations, which are applied to 3-D data sets for
DCT, and each block is transformed into a new signal. Follow-
decoration data on different bands. If the 3-D data set is
ing that, each block is represented by a number of frequency
hyperspectral image, KLT will remove the correlation between
coefficients that are arranged according to the frequency’s
image bands, building a more compressible data set.[9] KLT
magnitude. The discrete cosine transform (DCT) converts a
has shown a good performance in decorating lossless spectral
signal from spatial representation to frequency domain by
component.[10]
taking advantage of cosine functions. The DCT summarizes
sinusoids of various sizes and frequencies to represent an
image. Most of the visually relevant information about an IV. EXPERIMENTAL SETUP
image for a typical image is concentrated in just a few DCT lingkungan uji coba
coefficients, according to the property of DCT.[4]
V. PERFORMANCE EVALUATION
B. Discrete Wavelet Transform
VI. C ONCLUSION
Morlet and Grossman created the wavelet transform as an
R EFERENCES
alternative to older mathematical transformations to address
resolution issues in the 1980s. It is a mathematical function [1] A. Rizki, N. Ismail, and R. Mardiati, “Integrasi kriptografi algoritma
mars dan steganografi metode least significant bit (lsb) dengan media file
to describe data or functions. Continuous Wavelet Transform berekstensi *.wav,” TELKA - Telekomunikasi, Elektronika, Komputasi
(CWT) and Discrete Wavelet Transform (DWT) are the two dan Kontrol, vol. 3, pp. 57–67, 05 2017.
series that make up this transformation. The discovery of meth- [2] M. T. Banday and T. J. Shah, “A comparative study of image compres-
ods for decoding discrete time signals in 1976 served as the sion techniques,” no. October, 2013.
[3] K. H. Talukder and K. Harada, “Discrete wavelet transform for image
foundation for DWT [5]. The CWT analyzes signals using a compression and a model of parallel image compression scheme for
collection of fundamental functions coupled by straightforward formal verification.” in World Congress on Engineering. Citeseer, 2007,
scaling and transitions. While in DWT, digital filterization pp. 531–535.
[4] M. Ahuja and S. Shantaiya, “A review on image compression using dct
techniques are used to obtain the representation of the digital and dwt,” International Journal for Scientific Research & Development,
signal time scale. vol. 3, no. 10, 2015.
[5] D. Sripathi, “Efficient implementations of discrete wavelet transforms
using fpgas,” 2003.
C. Singular Value Decomposition [6] A. A. Adli, Rosal and Eko, “Image watermarking using triple transform
(dct-dwt-svd) to improve copyright protection performance,” 2018 Inter-
SVD is a mathematical approach to dividing images matrix national Seminar on Research of Information Technology and Intelligent
into three elements, namely one sigular element and two Systems (ISRITI), 2018.
unary elements. A watermark is usually embedded in a single [7] A. Touazi and M. Debyeche, “An svd-based scheme for mfcc com-
pression in distributed speech recognition system,” Automatic Speech
element to provide resilience and imperceptibility. Of the three Recognition and Understanding Workshop. December, 2013.
matrices produced from the SVD process, one of them has a [8] M. S. R. V.R.VijayKumar, A.Vidya, “Super resolution enhancement of
single value from the original matrix. What what is meant by medical image using quaternion wavelet transform with svd,” 017 4th
International Conference on Signal Processing, Communications and
a single value here is the distance between a matrix and single Networking (ICSCN -2017), March 16 – 18, 2017, Chennai, INDIA,
matrix. These single values Later it will be useful for matrices 2017.
JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2023 3

[9] C. Egho and T. Vladimirova, “Adaptive hyperspectral image com-

pression using the klt and integer klt algorithms,” 2014 NASA/ESA
Conference on Adaptive Hardware and Systems (AHS), 2014.
[10] N. R. M. Noor and T. Vladimirova, “Parallelised fault-tolerant integer
klt implementation for lossless hyperspectral image compression on
board satellites,” 2013 NASA/ESA Conference on Adaptive Hardware
and Systems (AHS-2013), 2013.