2022 OITS International Conference on Information Technology (OCIT)

Fractal Image Compression based on Discrete

Wavelet Transform
Pranjali Dwivedi Agya Mishra
Dept. of Electronics and Telecommunication Dept. of Electronics and Telecommunication
Engineering. Engineering.
Govt. Jabalpur Engineering College, Jabalpur,India Govt. Jabalpur Engineering College, Jabalpur,India
[email protected] [email protected]

Abstract—- Fractal Image compression is the type of data the domain and range pool and then by any partition
compression applied to the digital image to reduce their cost algorithm encoding is performed[11] and then
for storage and transmission. In this paper Fractal image performance parameter is calculated[1][9].
compression algorithm using a discrete wavelet transform
form has been projected. The concert of the (DWT) projected (B). Steps of FIC-
algorithm is estimated using compression ratio (CR). The
x Original Image
experimental result obtained from the projected algorithm
x
2022 OITS International Conference on Information Technology (OCIT) | 978-1-6654-9348-2/22/$31.00 ©2022 IEEE | DOI: 10.1109/OCIT56763.2022.00065

gives an improvement in terms of the CR. The Paper concludes Read Partition of the image into a non-overlapping
with a comparison with the existing algorithm and shows that Range block of m x m
the projected algorithm performs quite efficiently. x Selection of corresponding domain blocks of2m x
2m
Keywords—Fractal image compression (FIC), Discrete
wavelet transform (DWT), Compression Ratio (CR), Peak Signal x Apply DWT Discrete wavelet transform
to Noise Ratio (PSNR). The formula used is,
I. INTRODUCTION DWT = =

Image compression approaches the techniques which (1)

apply to digital images and reduce their cost for storage as
well as transmission[1]. This falls into two broad categories;
lossy compression techniques and lossless compression x In both domain & range apply the Quadtree
techniques[3]. Many compression techniques are used to partition scheme [11].
compress both images and videos. one of the most popular x Calculation of CR and PSNR [6].
methods is the fractal method[9]. FIC is a lossy type of
compression method of image compression which is based Compression Ratio = Original Image Data (2)
on the partitioning of an image that uses self similar concept Compressed Image
and has many different-different features like large encoding
time, fast decoding, better compression ratio, and PSNR = (3)
individuality of resolution [1-3]
Many projected algorithms, like algorithm using discrete (C). Discrete wavelet transform
cosine transform, using Huffman coding, using fractal
DWT is a transform that arises under the lossless nature
coding gives a better compression ratio CR and peak signal-
of image compression [5]. The below figure (Fig.1.)
to-noise ratio PSNR[3][5], whereas we know that the
compression ratio should be low for better image quality and represents the Discrete Wavelet Decomposition. This
PSNR should be range ranges between 20 to 25[5][7]. transforms the discrete-time signals into discrete wavelet
representations [5]. these transforms are mainly useful
In this research, a fractal image compression algorithm is for images to diminish unwanted and distorting noise.
projected using discrete wavelet transform DWT. The rest of DWT is a well-organized decomposition of the signal
the paper is systematized as: in Section 2, the theory of FIC, into lesser resolution and detailed resolution. Here, the
steps of FIC, discrete wavelet transform DWT[5], Quadtree ‘Haar-transform’ of discrete wavelet transform is used
partition scheme[7], and performance parameters CR and [5].
PSNR are described. In Section 3, the projected algorithm is
discussed and the flow chart of this algorithm is described. In III)II
Section 4, the experimental results are deliberated, where the I)II III)II
effectiveness of our projected algorithm is checked by III)III)IV III)II II
compression ratio CR[3] and finally in Section 5, the I)I
deduction of the proposed work is finished.
III)IV III)I
II. THEORY
(A). Fractal image compression
IV I
FIC is an algorithm, which divides the image into domain
and range blocks.in which, the domain block size is
multiplied by 2 then the range block size, and where the
range block comes into the center of the domain block
than by using any transform method image will split into Fig.1. Discrete Wavelet decomposition [5]

978-1-6654-9348-2/22/$31.00 ©2022 IEEE 307

DOI 10.1109/OCIT56763.2022.00065

Authorized licensed use limited to: UNIV OF ENGINEERING AND TECHNOLOGY TAXILA. Downloaded on May 10,2023 at 06:17:07 UTC from IEEE Xplore. Restrictions apply.
 (D). Partition scheme Which is used in the family or a living oh range model
The partition scheme is useful at the time of encoding, there are various partition scheme are avail lite-fixed
for the formation of range blocks. there are various size square block, Quadtree, Horizontal -vertical
partition schemes are available such as, fixed-size partition triangular partition in all these schemes we are
squared block, quadtree algorithm, horizontal-vertical applying the Quadtree algorithm scheme because this
partitioning, and triangular partitioning[7]. In all these scheme is based on the recursively splitting of image
schemes we are using the quadtree algorithm scheme quadrants which are chosen by us[7]. Then after the
because this scheme is based on the recursively splitting partition of the image looks like a tree in which the
of image quadrants[11]. which are chosen by us, here the array nonterminal mode has four descendants[11].
below figure (Fig.2)illustrates the Quadtree Partition Applying Quadtree algorithm for diagram encoding.
Scheme. then after partitions, the image looks like a tree here, figure (Fig.4) expresses the Flow chart of the
shape in which every non-terminal node has four compressed model
descendants[7].
FLOW CHART OF THE PROPOSED MODEL

Fig. 2. Quadtree Partition Scheme [5][7]

III. PROPOSED FIC BASED ON DWT MODEL

Fractal image compression is a lossy type compression
method for image compression it is a procedure that
divides the image into segments or two blocks[9]. There
are some steps which are of this algorithm, which are as
follows comes in the process [1][7].
1. Read the input image
2. Partition the image into non-overlapping range
blocks of size m x m. (See Fig.3)[6].
3. Again, partition the corresponding domain blocks of
size 2m x 2m. (See Fig.3)
4. Then the range looks like a range block sited at the
center of the domain blocks. (See Fig..3). Only a
square box should be used[6].
Domain Block
2m x 2m

Range Block

mxm
Fig. 4. Flow chart of compressed model

7. Calculate compression ratio and peak signal-to-

noise ratio.
Fig. .3. Domain Range Block[1] Compression ratio (CR)
5. Now, the term ‘transform’ comes we have to apply The ratio expresses how much the compressed data is
one transform in the range and domain block. than the original data reduction of the original row
image. (Eq No. 2)[5][8].
6. Now the time comes for encoding so here again we
have to choose one partition scheme for encoding[11].

308

Authorized licensed use limited to: UNIV OF ENGINEERING AND TECHNOLOGY TAXILA. Downloaded on May 10,2023 at 06:17:07 UTC from IEEE Xplore. Restrictions apply.
 Peak Signal to Noise Ratio (PSNR) –
It is an expression for the ratio between the maximum
possible value (or power) of a signal and the power of
distorting noise that affects the quality of its
representation[5][8].

PSNR =

Where

IV. EXPERIMENTAL RESULTS Fig.4 (c) Compressed Image

The projected algorithm is evaluated by two case studies Results of case study 1- 4(a),4(b),4(c).
there are two images considered in the case study each
TABLE 4.1RESULTS OF CASE STUDY 1
having any size or free size then in this algorithm, we
converted this size into 512 x 512 the experiment used Model Algorithm Compression Picture to
Matlab version R2021a. Ratio (CR) Noise Ratio
(PSNR)
4.1 Case study 1 Proposed FIC + DWT 1.86445 22.6793
Model
As input, we have taken the college image, which is
named the original photograph. Fig. 4)a) display the
original photograph, fig. 4)b) display the Quadtree 4.2 Case study2
photograph, and fig.4)c) display a compressed image
which is an output photograph. As input, we have taken another college photograph,
which is named the original photograph. Fig.4)d)
display the original photograph, Fig. 4)e) display the
Quadtree photograph and Fig.4)f) display a compressed
image which is an output photograph.

Fig.4 (a) Original photograph (dept. of E&TCE, JEC premises)

Fig.4 (d) Original photograph (dept. of E&TCE, JEC premises)

Fig.4 (b) Quadtree photograph

Fig.4 (e) Quadtree photograph

309

Authorized licensed use limited to: UNIV OF ENGINEERING AND TECHNOLOGY TAXILA. Downloaded on May 10,2023 at 06:17:07 UTC from IEEE Xplore. Restrictions apply.
 Here, figure (Fig.5) explains the plot of CR vs. PSNR (based
on the Existing model) and figure (Figure No.6) explains the
plot of CR vs PSNR (based on the projected model) is
shown.

Fig.4 (f) Compressed Image

Results of Case study 2- 4(d),4(e),4(f)

TABLE.4.2 RESULT OF CASE STUDY 2
Model Algorithm Compression Picture to
Ratio (CR) Noise Ratio
(PSNR)
Proposed Model FIC + DWT 1.9537 24.1211

4.3 Comparision of the projected algorithm and existing

algorithm
In this section, the proposed algorithm is compared with the Fig.6. CR vs. PSNR (Based on the proposed model)
existing algorithm in table 4.1 and graph 4.2 represents the Fig.6 accomplishes that the result obtained from the
comparison of both the existing and proposed model based proposed algorithm is better than the existing one.
on performance parameters CR and PSNR.
(V) CONCLUSION
TABLE 3 COMPARISON WITH THE EXISTING MODEL
The proposed fractal image compression-based algorithm
Model Algorithm Compression Picture to model has been fruitfully implemented. The present research
Ratio (CR) Noise Ratio
(PSNR) is found well-organized in improving the compression ratio
Existing Model FIC + DCT 7.97306 22.5442 and peak signal-to-noise ratio of a photograph. Research
trials were led for evaluating the performance of the
Proposed FIC + DWT 1.86445 22.6793 projected FIC model, which illustrates that it is quite better
Model
than the existing model. thus other FIC existing techniques
are having good performance[1][5] but in the proposed FIC
Table No.3 summarises that the result obtained from the model performance parameter is quite better and practically
projected algorithm model is better than the existing one. this model can be used anywhere for any type of photograph.

4.4 Figure in an experiment The future scope of our work is, these types of
compressed images can be useful in medical diagnoses like
X-ray and MRI images. Because it improves transmission
efficiency and prevents storage space or store data and also
takes very less transmission or processing time during
remote diagnosis.
ACKNOWLEDGMENT
I would like to thank our guide, Dr. Agya Mishra Madam,
for guiding this research. I would also like to thank my
friends and family who have supported me and provided me
with great insight into my research.
REFERENCES
[1] Jiji J S, Herlin L T, Ashwin G Singerji, and Sonal Wilson Pillai, “A
New Modified Fast Fractal Image Compression Algorithm in DCT
Domain,” Int. J. Eng. Res., vol. V6, no. 04, pp. 988–991, 2017, doi:
10.17577/ijertv6is040683.
[2] S. Liu, W. Bai, N. Zeng, and S. Wang, “A Fast Fractal Based
Compression for MRI Images,” IEEE Access, 2019, doi:
10.1109/ACCESS.2019.2916934.
Fig.5. CR vs PSNR (Based on Existing model) [1]

310

Authorized licensed use limited to: UNIV OF ENGINEERING AND TECHNOLOGY TAXILA. Downloaded on May 10,2023 at 06:17:07 UTC from IEEE Xplore. Restrictions apply.
 [3] S. Khatun and A. Iqbal, “A Review of Image Compression Using
Fractal Image Compression with Neural Network,” Int. J. Innov. Res.
Comput. Sci. Technol., 2018, doi: 10.21276/ijircst.2018.6.2.1.
[4] O. Svynchuk, O. Barabash, J. Nikodem, R. Kochan, and O. Laptiev,
“Image compression using fractal functions,” Fractal Fract., 2021,
doi: 10.3390/fractalfract5020031.
[5] R. Praisline Jasmi, B. Perumal, and M. Pallikonda Rajasekaran,
“Comparison of image compression techniques using Huffman
coding, DWT and fractal algorithm,” 2015, doi:
10.1109/ICCCI.2015.7218137.
[6] B. Senapatil and S. Kisan, “Color image compression using fractal
geometry,” 2018, doi: 10.1109/ICSNS.2018.8573638.
[7] G. Chauhan and A. Negi, “A review on fractal image compression,”
Int. J. Sci. Technol. Res., vol. 8, no. 12, pp. 1893–1899, 2019, doi:
10.5120/14830-3081.
[8] A. M. H. Y. Saad, M. Z. Abdullah, A. M. Alduais Nayef, and A. S. H.
Abdul-Qawy, “An Improved Full-search Fractal Image Compression
Method with Dynamic Search Approach,” 2020, doi:
10.1109/ICCSCE50387.2020.9204926.
[9] V. Drakopoulos, “Fractal-Based Image Encoding And Compression
Techniques,” . January, pp. 48–55, 2013.
[10] B. Mohammed Ismail, T. B. Reddy, and B. E. Reddy, “Spiral
Architecture-based Bybrid Fractal Image Compression,” 2017, doi:
10.1109/ICEECCOT.2016.7955179.
[11] Chandran, Saravanan &Vrindavanam, Jayavrinda & Mahanti,G. “A
Survey of Image Compression Methods” International Conference &
Workshop on Recent Trends in Technology, (TCET) 2012.

311

Authorized licensed use limited to: UNIV OF ENGINEERING AND TECHNOLOGY TAXILA. Downloaded on May 10,2023 at 06:17:07 UTC from IEEE Xplore. Restrictions apply.