An Improved Image Compression Technique Using

Huffman Coding and FFTS

Ishan Agarwal
Electronics and Communication Engineering Department
National Institute of Technology,Hamirpur
Hamirpur, Himachal Pradesh- 177005
194540@nith.ac.in

Abstract—The two fundamental methods for compressing im-

ages are Huffman coding and the Fourier Transform approach.
When compared to Huffman coding, the Fourier transform
approach is far more powerful. apply to multi orders. The
goal of this work is to examine the performance of Huffman
coding and Fast Fourier Transform in the compression of digital
photographs. Matlab tool is necessary for parameter calculations.
These techniques are compared in terms of various parameters,
including mean square error (MSE), peak signal-to-noise ratio
Fig. 1. Image compression process
(PSNR), compression ratio (CR), and bits per pixel (BPP), for
a variety of input images of different sizes. It also involves a
new method of splitting or dividing an input image into equal
rows and columns and, at the conclusion, summing up all the image compression techniques: lossy compression and lossless
compressed individual images. method [2]. Initially, the coding procedure gathers all the prob-
I. I NTRODUCTION ability for specific motifs in decreasing order [4]. Beginning at
the very beginning, a tree. Symbols can be found on every leaf
Image compression is a highly helpful tool for conserving
of the tree. Steps are taken in the procedure. For each Step one
storage space. The primary goal of compression is to reduce
compares two probabilities and chooses the lowest one. chosen
the amount of unnecessary visual data in order to improve
likelihood to the tree’s top is added. Removed a particular
the efficiency of data transmission or storage. Compression,
likelihood from the list. It continues. until each probability
which simply increases the amount of memory needed without
has been chosen, if necessary. The Huffman approach used in
impacting image quality, is the process of reducing or deleting
this paper will be completed or used following the image’s
unneeded information or data from an image [1]. The only way
breakdown into smaller components (rows and Applying this
to determine which picture compression approach is superior
method to each portion (columns), if necessary [3].
is to calculate the processing time, or compression speed,
which depends on the processor’s speed. Bits from images are III. FAST F OURIER T RANSFORM
minimised by the use of image compression [2]. The purpose The Fourier Transform is a mathematical method used to
of compression is to remove as few superfluous bits from a transform signals from one domain to another. Inverse Fourier
picture as possible so that it may be stored or shared with Transform transforms frequency domain signals back into
maximum efficiency. Two types of image compression exist. time domain signals, while Fourier Transform transforms time
The first one is lossy, while the second is lossless. For artificial domain signals into frequency domain signals [5]. Discrete
images like technical drawings, icons, or comics, lossless Fourier Transform (DFT) is developed into FFT [5, 6]. It
compression is used. Lossy compression use a low bit rate. completes equivalent tasks more quickly and more effectively
For image scans generated for archive purposes, the lossless [7, 8]. Because it transforms a complex discrete time signal
compression [1] approach may be utilised. When dealing into a simple discrete time frequency signal, DFT is strongly
with natural images, such as photos in applications, lossy advised for frequency studies [7]. The discrete nature of the
approaches are helpful. Compression is another application FFT can be easily examined using Matlab or in real-time.
for Fourier Transform. In Fig. 1 [3], a block schematic of the N-point DFT is simple to calculate using the FFT, with a
compression process is displayed. Digital photos have better complexity of O(N log N) [9].
noise performance thanks to a process called quantization,
which reduces noise.
II. H UFFMAN C ODING A LGORITHM
Data types are not relevant when using the Huffman coding
method for data compression. There are two main types of
 D. Bits per Pixel (BPP)
It is given as-
V. M ATHEMATICAL A NALYSIS
The following compact matrix form can be used to represent
the entire MxM digital image, where t(r, s) denotes the digital
image and each element of this matrix array is referred to as
an image pixel.

Fig. 2. Flow chart of the technique used

IV. I MAGE Q UALITY PARAMETERS

There are four major important parameters measured be-
tween uncompressed image and compressed image, these are
following- This digital image t(r,s) is divided into a set of two-row
A. Compression Ratio (CR) and two-column, non-overlapping four sub-images. this can
be illustrated by-
It is essentially described as the ratio between the sizes of
compressed and uncompressed images [1]. It is provided as:

B. Mean Square Error (MSE)

VI. R ESULT ANALYSIS
The difference between an uncompressed and compressed In order to conceptually differentiate between Huffman and
image squared is what it is known as. It is provided as: Where FFT, the author calculates various image quality parameters
in this paper, including CR, PSNR, MSE, and BPP. By
analysing the above-mentioned parameters, the author can then
determine which will produce the best results. These factors
are determined for 256x256, 512x512 of three images and the
results are shows in the tables to the corresponding image.
A. Original images
R is row and C is column h’(a,b), the function of is compressed
image and h(a, b) is the function of uncompressed image. B. Compressed images
The compressed image can be obtain by adding these
C. Peak Signal to Noise Ratio (PSNR) t 1(r,s), t 2(r,s), t 3(r,s), t 4(r,s) matrixes, it can be given as
PSNR stands for the amount of noise an output signal By Huffman Coding :-
contains. It is written in logarithmic form in mathematics. By FFT,
Image size and PSNR are correlated in a direct manner.
 Fig. 3. Original image

Fig. 8. Output image of 512x512

Fig. 4. Compressed image

Fig. 5. Original image Fig. 9. PSNR for FFT and HUFFMAN

Fig. 6. Output image of 256x256

VII. C ONCLUSIONS
Although the new technique of dividing an image into
various rows and columns will aid with image transmission
security, it will also produce better results in image com-
pression. Following Matlab analysis of the various image
properties, it was determined that both compression techniques
would yield good results, but that the FFT technique would
outperform the Huffman strategy as seen by the tables of
image quality parameters. Therefore, it is obvious that the FFT
Fig. 7. Output image of 512x512 technique for compression outperforms the Huffman technique
for compressing images.
 ACKNOWLEDGMENT
I’d like to thank Prof. Sandeep Kumar Singh, my mentor,
for providing me with this incredible opportunity to work on
the topic ”An Improved Image Compression Technique Using
Huffman Coding and FFTS.” Working on this project taught
me a lot about Huffman coding and image compression, as
well as a lot about the various tools in general. Second, I am
grateful for my parents’ and friends’ assistance in completing
the project in such a short time. Finally, I’d want to express
my gratitude to my seniors for their technical and conceptual
assistance and support.
R EFERENCES
[1] Chen, C.C.: On the selection of image compression algorithms. In:
Pattern Recognition, Brisbane, Australia, August 1998
[2] Van loan, C.: Transform computational framework for fast Fourier. In:
SIAM (1992)
[3] Yusong, Y., Guangda, S., Chunmei, W., Qingyum, S.: Invertible integer
FFT applied on lossless image compression. IEEE Robot. Intell. Syst.
Signal Process. 2, 1219–1223 (2003)
[4] Zayed, A.I.: On the relationship between the Fourier and fractional
Fourier transforms. IEEE Signal Process. Lett. 3, 310–311 (1996)
[5] Hu, T. C.; Tucker, A. C. (1971). ”Optimal Computer Search Trees
and Variable-Length Alphabetical Codes”. SIAM Journal on Applied
Mathematics. 21 (4): 514. doi:10.1137/0121057
[6] Gribov, Alexander (2017-04-10). ”Optimal Compression of a Polyline
with Segments and Arcs”. arXiv:1604.0747
[7] Kaushik, C.S.H., Elamaran, V.: A tutorial review on discrete Fourier
transform with data compression application. In: ICGCCEE, pp. 1–6,
March 2014
[8] Sahnoun, K., Benabadji, N.: On-board satellite image compression using
the Fourier transform and Huffman coding. In: WCCIT, pp. 1–5, June
2013
[9] Jain, A. k.: Image data compression: a review. In: Proceedings of IEEE,
vol. 69, pp. 349–389 (1981)
[10] Gonzalea, R.C., Woods, R.E.: Digital Image Processing, 2nd edn.
Prentice Hall, Upper Saddle River (2004)
[11] Abrahams, J. (1997-06-11). Written at Arlington, VA, USA. Division
of Mathematics, Computer Information Sciences, Office of Naval
Research (ONR). ”Code and Parse Trees for Lossless Source Encoding”.
Compression and Complexity of Sequences 1997 Proceedings. Salerno:
IEEE: 145–171. CiteSeerX 10.1.1.589.472