An Algorithm for Improving the Quality of Compacted JPEG Image by Minimizes the Blocking Artifacts

International Journal of Computer Graphics & Animation (IJCGA) Vol.2, No.2/3, July 2012
DOI : 10.5121/ijcga.2012.2202 17
AN ALGORITHM FOR IMPROVING THE QUALITY OF
COMPACTED JPEG IMAGE BY MINIMIZES THE
BLOCKING ARTIFACTS
Sukhpal Singh1
1
Department of Computer Science and Engineering, Thapar University, Patiala, India.
ssgill@hotmail.co.in
ABSTRACT
The Block Transform Coded, JPEG- a lossy image compression format has been used to keep storage and
bandwidth requirements of digital image at practical levels. However, JPEG compression schemes may
exhibit unwanted image artifacts to appear - such as the ‘blocky’ artifact found in smooth/monotone areas
of an image, caused by the coarse quantization of DCT coefficients. A number of image filtering
approaches have been analyzed in literature incorporating value-averaging filters in order to smooth out
the discontinuities that appear across DCT block boundaries. Although some of these approaches are able
to decrease the severity of these unwanted artifacts to some extent, other approaches have certain
limitations that cause excessive blurring to high-contrast edges in the image. The image deblocking
algorithm presented in this paper aims to filter the blocked boundaries. This is accomplished by employing
smoothening, detection of blocked edges and then filtering the difference between the pixels containing the
blocked edge. The deblocking algorithm presented has been successful in reducing blocky artifacts in an
image and therefore increases the subjective as well as objective quality of the reconstructed image.
KEYWORDS
Image processing, compression, blocking artifacts, DCT.
1. INTRODUCTION
As the usage of computers continue to grow, so too does our need for efficient ways for storing
large amounts of data (images). For example, someone with a web page or online catalog – that
uses dozens or perhaps hundreds of images-will more likely need to use some form of image
compression to store those images. This is because the amount of space required for storing
unadulterated images can be prohibitively large in terms of cost. Several methods for image
compression are available today and these are categorized as: lossless and lossy image
compression. The JPEG is a widely used form of lossy image compression standard that centers
on the Discrete Cosine Transform (DCT). The DCT works by separating images into parts of
differing frequencies. During a step called quantization, where part of compression actually
occurs, the less important frequencies are discarded. Then, only the important frequencies remain
and are used to retrieve the image in the decompression process. As a result, the reconstructed
images contain some distortions. At low bit-rate or quality, the distortion called blocking artifact
is unacceptable [1]. This paper work deals with reducing the extent of blocking artifacts in order
to enhance the both subjective as well as objective quality of the decompressed image.

18
1.1 Motivation
JPEG defines a "baseline" lossy algorithm, plus optional extensions for progressive and
hierarchical coding. Most currently available JPEG hardware and software handles only the
baseline mode. It contains a rich set of capabilities that make it suitable for a wide range of
applications involving image compression. JPEG requires little buffering and can be efficiently
implemented to provide the required processing speed for most cases. Best Known lossless
compression methods can compress data about 2:1 on average. Baseline JPEG (color images at 24
bpp) can typically achieve 10:1 to 20:1 compression without visible loss and 30:1 to 50:1
compression visible with small to moderate defects. For Gray Images (at 8 bpp), the threshold for
visible loss is often around 5:1 compression. The baseline JPEG coder is preferable over other
standards because of its low complexity, efficient utilization of memory and reasonable coding
efficiency. Although more efficient compression schemes do exist, but JPEG is being used for a
long period of time that it has spread its artifacts over all the digital images [6]. The need for a
blocking artifact removal technique is therefore a motive that constantly drives new ideas and
implementations in this field. Considering the wide spread acceptance of JPEG(baseline) standard
[16], this paper suggested a post processing algorithm that does not make any amendments into
the existing standard, and reduces the extent of blocking artifacts. That is, the work is being done
to improve the quality of the image.
1.2 Paper Outline
This paper deals with three of the image processing operations:
1.2.1 Image restoration- Restoration [12] takes a corrupted image and attempts to recreate a
clean original. It is clearly explained in the figure 1.1.
Figure 1.1a) Original image b) Image after restoration
1.2.2 Image Enhancement - Image Enhancement alters an image to makes its meaning clearer
to human observers [13]. It is often used to increase the contrast in images that are overly
dark or light explained in the figure 1.2.
Figure 1.2 a) Original image b) Image after enhancement

19
1.2.3 Image Compression- Image compression is the process that helps to represent image data
with as few bits as possible through exploiting redundancies in the data while maintaining an
appropriate level of quality for the user [14]. Lossy compression is shown in figure-1.3.
Figure 1.3 a) Original image b) Image after compression
1.3 Image Compression
As the beginning of the third millennium approaches, the status of the human civilization is best
characterized by the term “Information Age” [5]. Information, the substance of this new world,
despite its physical non-existence can dramatically change human lives. In a sense, being in the
right place at the right time depends on having the right information, rather than just having luck.
Information is often stored and transmitted as digital data [6]. However, the same information can
be described by different datasets in figure 1.4.
DATA = REDUNDANT DATA + INFORMATION
Figure-1.4: Relationship between data and information [6]
The shorter the data description, usually the better, since people are interested in the information
and not in the data. Compression is the process of transforming the data description into a more
succinct and condensed form. Thus improves the storage efficiency, communication speed, and
security [9]. Compressing an image is significantly different than compressing raw binary data.
This is because images have certain statistical properties which can be exploited by encoders
specifically designed for them.
1.4 Motivation behind image compression [11]
A common characteristic of most images is that the neighboring pixels are correlated and
therefore contain redundant information [20]. The foremost task then is to find less correlated
representation of the image. In general, three types of redundancy can be identified:
Redundant Data
Information

20
1.4.1 Coding Redundancy: If the gray levels of an image are coded in a way that uses more code
symbols than absolutely necessary to represent each gray level, the resulting image is said to have
coding redundancy.
1.4.2 Interpixel Redundancy: This redundancy is directly related to the interpixel correlations
within an image. Because the value of any given pixel can be reasonably predicted from the value
of its neighbors, the information carried by individual pixels is relatively small. Much of the
visual contribution of a single pixel to an image is reduntant; it could have been guessed on the
basis of the values of its neighbors.
1.4.3 Psychvisual Redundancy: This redundancy is fundamentally different from other
redundancies. It is associated with real or quantifiable visual information [21]. Its elimination is
possible only because the information itself is not essential for normal visual processing. Since
the elimination of psychvisually redundant data results in a loss of quantitative information, it is
commonly referred to as quantization.
1.5 Image compression model
Figure-1.5: Image compression Model [16]
As the above figure shows, a compression system consists of two distinct structural blocks: an
encoder and a decoder. An input image f(x,y) is fed into the encoder, which creates a set of
symbols from the input data. After transmission over the channel, the encoded representation is
fed to the decoder, where the reconstructed output image f’(x,y) is generated. In general, f’(x,y)
may or may not be the exact replica of f(x,y).The encoder is made up of a source encoder, which
removes input redundancies, and a channel encoder, which increases the noise immunity of the
source encoder’s output. Same is in the case of decoder, but functions in reverse direction
explained in figure 1.5.
1.5 Compression Techniques
There are two different ways to compress images-lossless and lossy compression.
1.6.1 Lossless Image Compression
A lossless technique means that the restored data file is identical to the original explained in
figure 1.6. This type of compression technique is used where the loss of information is
unacceptable [22]. Here, subjective as well as objective qualities are given importance. In a
nutshell, decompressed image is exactly same as the original image.
Source
Encoder
Channel
Encoder
Channel
Source
Decoder Channel
f(x,y)
f’(x,y)

21
a). Original Image b).Decompressed Image
Figure-1.6: Relationship between input and output of Lossless Compression
1.6.2 Lossy Image Compression
It is based on the concept that all real world measurements inherently contain a certain amount
of noise. If the changes made to these images, resemble a small amount of additional noise, no
harm is done [23]. Compression techniques that allow this type of degradation are called lossy
explained in figure 1.7. The higher the compression ratio, the more noise added to the data. In a
nutshell, decompressed image is as close to the original as we wish.
a).Original Image b).Decompressed Image
Figure-1.7: Relationship between input and output of Lossy Compression
Lossless compression technique is reversible in nature, whereas lossy technique is irreversible.
This is due to the fact that the encoder of lossy compression consists of quantization block in its
encoding procedure.
1.6 JPEG
JPEG (pronounced "jay-peg") is a standardized image compression mechanism. JPEG also stands
for Joint Photographic Experts Group, the original name of the committee that wrote the standard.
JPEG is designed for compressing full-color or gray-scale images of natural, real-world scenes. It
works well on photographs, naturalistic artwork, and similar material [18]. There are lossless
image compression algorithms, but JPEG achieves much greater compression than with other
lossless methods. JPEG involves lossy compression through quantization that reduces the number
of bits per sample or entirely discards some of the samples. The usage of JPEG compression
method is motivated because of following reasons:-
1.7.1 The compression ratio of lossless methods is not high enough for image and video
compression.
1.7.2 JPEG uses transform coding, it is largely based on the following observations:

22
Observation 1: A large majority of useful image contents change relatively slowly across images,
i.e., it is unusual for intensity values to alter up and down several times in a small area, for
example, within an 8 x 8 image block [24].
Observation 2: Generally, lower spatial frequency components contain more information than the
high frequency components which often correspond to less useful details and noises [26].
Thus, JPEG is designed to exploit known limitations of the human eye, notably the fact that small
color changes are perceived less accurately than small changes in brightness. JPEG can vary the
degree of lossiness by adjusting compression parameters [22]. Useful JPEG compression ratios
are typically in the range of about 10:1 to 20:1. Because of the mentioned plus points, JPEG has
become the practical standard for storing realistic still images using lossy compression.JPEG
(encoding) works as shown in the figure 1.8. The decoder works in the reverse direction. As
quantization block is irreversible in nature, therefore it is not included in the decoding phase. It is
clearly explained in the figure 1.8.
Figure 1.8: Steps in JPEG Compression [22]
A major drawback of JPEG (DCT-based) is that blocky artifacts appear at low bit-rates in the
decompressed images. Such artifacts are demonstrated as artificial discontinuities between
adjacent image blocks. An image illustrating such blocky artifacts is shown in the figure below:-
Figure1.9: a).Actual Image b).Blocked Image

23
2. LITERATURE SURVEY
In practice, it is always a tradeoff between the coding bit rate and the coded image quality.
Generally speaking, increasing coding bit rate can improve the quality of the reconstructed image,
but it is limited by channel bandwidth or storage capacity. In order to achieve specified bit rate, a
large quantization step for transformed coefficients should be adopted, so the blocking artifacts
are more obvious at high compressed ratios, which often limit the maximum compression
capacity that can achieve (Wang, 2004). Many algorithms have been proposed for reducing these
blocking artifacts. These algorithms can be classified into two major categories. One is to use
different encoding schemes, such as the interleaved block transform, the lapped transform, and
the combined transform. Further, post-processing algorithms can be categorized as spatial,
frequency (DCT-based) and combined (hybrid) algorithms based on the domain in which they
operate. The research done by various authors in the same field have been summarized in Table
2.1.
Table 2.1: Summary of research done by different authors
Sr. Author’s
name
Paper Title Focus
1. Ramamurthi
and Gersho
space-variant filter
that adapts to local
characteristics of the
signal
The algorithm distinguishes edge pixels from
non-edge pixels via a neighborhood testing and
then switches between a 1D filter and a 2D filter
accordingly to reduce blocking effects [11].
2. Long Adaptive Deblocking
of Images with DCT
Compression
Proposed a very unique concept of “noise
injection.” By adding a small amount of random
noise to a blocky image in the DCT domain (the
“noise injection” process), he converted the
blocky image into an image more like being
corrupted with “Gaussian-type” noise, thus
changed the image deblocking problem into a
denoising problem [3, 24].
3. Robertson
DCT Quantization
Noise in Compressed
Images
Provides a spatial domain model of the
quantization error based on statistical noise
model of the error introduced when quantizing
the DCT coefficients [4, 22].
4. Tuan
Blocking artifacts
removal by a hybrid
filter method
Suggested a method that simultaneously
performs an edge-preserving and a low-pass
filtering of the degraded image [19].
5. Aria
Enhancement of
JPEG-Compressed
Images by Re-
application of JPEG
Simply re-applies JPEG to the shifted versions of
the already-compressed image, and forms an
average [2].
6. Kwan
Blocking Artifacts
Reduction Algorithm
in Block Boundary
area using Neural
Network
Proposed an algorithm using block classification
and feed forward neural network filters in the
spatial domain [12].
7. Park Blocking Artifacts
Reduction in Block-
Coded Images Using
Self-Similarity
Proposed algorithm considering piecewise self-
similarity within different parts of the image as a
priori to give reasonable modification to the
block boundary pixels [17].

24
8. Gothundar
umun
Total Variation for the
Removal of Blocking
Effects in DCT Based
Encoding
Proposed a non-linear method that reduces
artifacts by minimizing the total variation via a
level set formulation [7].
9. Wenfeng
A De-Blocking
Algorithm and a
Blockiness Metric for
Highly Compressed
Images
Proposed a de-blocking algorithm based on the
number of connected blocks in a relatively
homogeneous region, the magnitude of abrupt
changes between neighboring blocks, and the
quantization step size of DCT coefficients [3, 26].
10. Averbuch
Deblocking of Block-
Transform
Compressed Images
Using Weighted Sums
of Symmetrically
Aligned Pixels
Appllied weighted sums on pixel quartets, which
are symmetrically aligned with respect to block
boundaries [3].
11. Yang
Regularized
Reconstruction to
Reduce Blocking
Artifacts of Block
Discrete Cosine
Transform
Compressed Images
Suggested two methods for solving this
regularized problem. The first was based on the
theory of projections onto convex sets (POCS)
while the second was based on the constrained
least squares (CLS) approach [28]. For the POCS-
based method, a new constraint set was defined
that conveys smoothness information not captured
by the transmitted B-DCT coefficients [27], and
the projection onto it was computed. For the CLS
method an objective function was proposed that
captures the smoothness properties of the original
image [4].
12. Triantafyllidis
Blockiness Reduction
in JPEG Coded
Images
Suggested an algorithm that firstly estimates the
AC coefficients based on their observed
probability distribution and then, a postprocessing
scheme consisting of a region classification
algorithm and a spatial adaptive filtering is applied
for blockiness removal [9, 22].
13. Hyuk
Blocking-Artifact
Reduction in Block-
Coded Images Using
Wavelet-Based Subband
Decomposition
Proposed a post-processing method in the
wavelet transform domain. Knowing that sub
band coding does not suffer from blocky noise,
the proposed technique is designed to work in the
sub band domain [8].
14. Triantafyllidis
Detection of Blocking
Artifacts of
Compressed Still
Images
Detects the regions of visible blocking artifacts
and uses the estimated relative quantization error
calculated when the DCT coefficients are modeled
by a Laplacian probability function [18].
15. Zhao
Postprocessing
technique for blocking
artifacts reduction in
DCT domain
Stated that the DCT distributions of differently
shifted blocks before quantization are
approximately identical, whereas the DCT
distributions of differently shifted blocks in the
compressed image are considerably different and
used the difference to expose the blocking
artifacts [18].
16. Wang
Adaptive Reduction of
Blocking Artifacts in
DCT Domain for
Removes these discontinuities by Walsh
transform and local threshold technology, the
precision of edge detection by Sobel operator is

25
Highly Compressed
Images
improved and blurring in objective edge is
reduced [20-21].
17. Alan’s
Blocking Artifacts
Suppression in Block-
Coded Images Using
Overcomplete Wavelet
Representation
Suggested a non-iterative, wavelet-based
deblocking algorithm exploiting the fact that
block discontinuities are constrained by the dc
quantization interval of the quantization table [4],
as well as the behavior of wavelet modulus
maxima evolution across wavelet scales to derive
appropriate threshold maps at different wavelet
scales [13].
18. Kizewski’s
Image Deblocking
Using Local
Segmentation
Suggested an image deblocking filter by
employing local segmentation techniques prior to
applying a simple value averaging model to a
local neighborhood of 3x3 DCT coefficients
[11].
19. Wang
Fast Edge-Preserved
Postprocessing for
Compressed Images
Proposed a fast algorithm based on the concept
to decompose a row or column image vector to a
gradually changed signal and a fast variation
signal [25].
20. Nallaperumal
Removal of blocking
artifacts in JPEG
compressed images
using Dual Tree
Complex Wavelet
Filters for Multimedia
on the Web
Proposed an algorithm based on the fact that the
high frequency [15] details of the coded image
are mainly contaminated by quantization noise
[10].
21. Luo
Removing the
Blocking Artifacts of
Block-Based DCT
Compressed Images
Proposed an adaptive approach. For smooth
regions, the method takes advantage of the fact
that the original pixel levels in the same block
provide continuity and use this property and the
correlation between the neighboring blocks to
reduce the discontinuity of the pixels across the
boundaries [14].
3. PRESENT WORK
3.1 Problem Formulation
In JPEG (DCT based) compresses image data by representing the original image with a small
number of transform coefficients. It exploits the fact that for typical images a large amount of
signal energy is concentrated in a small number of coefficients. The goal of DCT transform
coding is to minimize the number of retained transform coefficients while keeping distortion at an
acceptable level.In JPEG; it is done in 8X8 non overlapping blocks. It divides an image into
blocks of equal size and processes each block independently. Block processing allows the coder
to adapt to the local image statistics, exploit the correlation present among neighboring image
pixels, and to reduce computational and storage requirements. One of the most degradation of the
block transform coding is the “blocking artifact”. These artifacts appear as a regular pattern of
visible block boundaries. This degradation is a direct result of the coarse quantization of the
coefficients and the independent processing of the blocks which does not take into account the
existing correlations among adjacent block pixels. In this paper attempt is being made to reduce
the blocking artifact introduced by the Block DCT Transform in JPEG.

26
3.2. Objective
The primary objective of this paper is to develop a post-processing algorithm for the reduction of
blocking artifacts, as it does not need to modify the existing standard. The algorithm tried to
fulfill the following criteria’s too.
3.2.1. Reduce the extent of blocking artifacts
3.2.2. Make efficient utilization of resources
3.2.3. Preserves the edges.
3.2.4. Do not result in other type of artifacts
3.3. Design and Implementation
3.3.1 Algorithm
This algorithm is implemented in MatLab 7.There are mainly three segments:
a) Segment-1
From the earlier discussion, it has been cleared that in JPEG the image data is dealt block by
block basis. That is DCT is applied on each block independently and the same for the
quantization .Because of this fact, blocking artifacts came into existence representing the lossy
nature of the JPEG standard. This fact is exploited in this segment. As difference exists
between each block, an attempt is made (spatially) to smooth out this difference.
There are many smoothing filters that can easily smoothes the difference, but it results in
blurring (which is another artifact).The smoothening is being done in a very tactful manner.
Due to low quality, edges are quite visible in the image (blocked). Smoothening is done by
gradually decreasing the difference between the block edges. The point is also considered that
the blocking artifact should not move inside the block, that’s why gradual attempt (that is
neighboring pixels are also manipulated) is made to reduce the difference. The key idea behind
smoothing the blocked image is to reduce the extent of blockiness without blurring the image.
Smoothening is done by considering six neighboring pixels on either side of pixels containing
the block edge. The algorithm used in this segment is described below:
Input: Preprocessed Image
Output: Uniformly deblocked Image
Assumptions: a(f(x,y)) and b(f(x,y+1)) are two adjacent pixels having block boundary in
between, .in a vertical direction.
Algorithm:
If |a-b|<threshold
//no change in the pixels values.
Else
s = |a-b|/2
if a<b
f(x,y)=f(x,y)+s,f(x,y+1)-s
f(x,y-1)=f(x,y-1)+s/2, f(x,y+2)=f(x,y+2)-s/2
f(x,y-2)=f(x,y-2)+s/2, f(x,y+3)=f(x,y+3)-s/4
elseif a>b
f(x,y)=f(x,y)-s,f(x,y+1)+s

27
f(x,y-1)=f(x,y-1)-s/2, f(x,y+2)=f(x,y+2)+s/2
f(x,y-2)=f(x,y-2)-s/2, f(x,y+3)=f(x,y+3)+s/4
else
//nothing
end
The algorithm is repeated for the all the eight pixels along the vertical edge and similar
algorithm is used for the horizontal edges.
b) Segment-2
As mentioned, an attempt is made in the previous segment to reduce the blocking artifacts and
the artifacts are reduced to some extent. But in the previous segment, blocking artifacts are
dealt in an uniform manner. In this segment, the blocked edges are first detected and then
deblocked by using Gaussian formula. Here detection and deblocking are interlinked. That is,
the pixel being deblocked depends upon the detection criteria.
BLOCK EDGE
L3 L1 R1 R3
L4 L2 G0 R2 R4
Figure-3.1: Detection Sequence
Input: Eight rows of ten pixels having the edge in the centre, which is being detected
against blockiness.
Output: Deblocked Edge
Algorithm:
1. Initialize counter = 0.
2. Considering first row.
3. Assign G0= |x0-y7| //difference between two boundary pixels.
4. Assign the difference between each pair of adjacent pixels on left and right-hand side
of the block boundary are also calculated and denoted by Li and Ri (i=1,2,3,4)
respectively.
1. If MAX(L1,L2,L3,L4)<G0 (1) or MAX(R1,R2,R3,R4)<G0 (2)
// boundary gap is detected
//current row is marked
counter= counter+1 // increment the counter
2. Repeat the steps 3-5 for the rest seven rows.
3. If counter> TH, then blocking artifact is claimed // TH threshold value
4. 1-D filter using Gaussian formula is applied to {x0,y7.y6}(if equation 1 holds)
or
{y7,x0,x1} (if equation 2 holds) having window size=5, along the marked rows.
Detection is done by following the above mentioned detection algorithm for each
internal edge of the image and if these follow the decided criteria, only then the edge
x4
x3
x2
x1
x0
y7
y6
y5
y4
y3

International Journal of Computer Graphics Animation (IJCGA) Vol.2, No.2/3, July 2012
28
is marked as BLOCKED edge. The blocked edge is deblocked by applying gaussain
formula:
Y = ∑N
j=1 xjwj / ( ∑N
j=1 wj )
where wj=exp[-(xc-xj)2
/2ε2
]
N is the window size
XC is the centre pixel
ε = average difference from centre pixel
c) Segment-3
Although there is no consensus on a method to measure block artifacts in images, one measure
has been used in most of the papers encountered - Peak Signal to Noise Ratio (PSNR). PSNR
is basically a logarithmic scale of the mean squared difference between two sets of values
(pixel values, in this case). It is used as a general measure of image quality, but it does not
specifically measure blocking artifacts. In observed literature, PSNR is used as a source-
dependant artifact measure, requiring the original, uncompressed image to compare with.
PSNR is defined as:
PSNR = 20log10 (255/√MSE)
Where MSE = ∑ (Bi − Ai)/n
where i = 0 to n and n is the number of pixels in the image.
It is easily seen that this blockiness measure is not actually aware of the artifacts it is
measuring - it is simply a gauge of how different the corresponding (that is, the same position)
pixel values are between two images. Because a blocky image is different from the original
and a severely blocky image is more, so PSNR is an acceptable measure, and hence the
primary measure used to compare the proposed method. However, two images with
completely different levels of perceived blockiness may have almost identical PSNR values.
The working of the algorithm is explained by the flowchart drawn in figure 3.2.
Figure: 3.2. Flowchart of the Proposed Algorithm
3.4 Test Images
Seven images have been selected for checking the validation of the proposed algorithm. All are
512X512 .jpg images at different quality factors. It is clearly explained in the figure 3.3 and
figure 3.4.
Start Enter the Blocked Pre-process the Image
Block Edge
Uniform
Blocked Edge
Detection based Evaluation Stop

29
Pentagon(Q=100) Pentagon(Q=10)
Pentagon( Q=5) Pentagon( Q=1)
Figure-3.3: Pentagon Image (512X512)
Bridege(Q=100) Bridge(Q=10)
Bridge(Q=5) Bridge(Q=1)
Figure-3.4: Bridge Image (512X512)
4. RESULTS AND DISCUSSIONS
The proposed algorithm is tested on various images with different characteristics. The algorithm
is applied on seven images (as mentioned in the previous chapter) at different quality parameters
(Q=10, 5, and 1).The results are shown in Fig: 4.1, 4.2, 4.3 and Table-1.

30
Proposed Method
JPEG Method
Figure: 4.1 Relationships between PSNR and Quality of Elaine Image
The above figure shows the relationship between PSNR and Quality of Elaine Image using JPEG
Method (green) and Proposed Method (red).It is very clear from the plot that there is increase in
PSNR value of image with the use of proposed method over the JPEG method. This increase
represents improvement in the objective quality of the image.
Proposed Method
JPEG Method
Figure: 4.2 Relationships between MSE and Quality of Elaine Image
The above figure shows the relationship between MSE and Quality of Elaine Image using JPEG
Method (green) and Proposed Method (red).It is very clear from the plot that there is decrease in
MSE value of image with the use of proposed method over the JPEG method. This decrease
represents improvement in the objective quality of the image.

31
Baboon Barbara
Elaine Bridge
Figure: 4.3 Relationships between PSNR and Quality
Figure 4.3 shows the relationship between PSNR and quality factor of various images using the
proposed method. The top-most line (green) representing Elaine image shows that at quality=1,
PSNR= 31.7557; at quality=5, PSNR= 33.4897 and at quality=10, PSNR=35.2924.This shows
that PSNR value increases with increase in quality of image. This reveals that the relationship
between quality and PSNR value of Lena image obtained by the proposed method is in consent
with the fact that PSNR value increases with increase in the quality of image. The same fact
fulfilled by other images represents the validity of the proposed algorithm. The application of the
proposed algorithm on Lena Image (512X512) at different bpp is shown in figure: 4.4-4.6.
At Quality Parameter = 10(0.25 bpp)
Figure-4.4: a). Blocked Image Figure-4.4: b) Deblocked Image
Figure-4.5: a). Blocked Image Figure-4.5: b). Deblocked Image

32
Figure-4.6: a). Blocked Image Figure- 4.6: b). Deblocked Image
The figure 4.4-4.6 shows the improvement in deblocked image after application of proposed
algorithm to the blocked image at various quality (bpp) values. This represents that the proposed
algorithm increases the subjective quality of the image. The table below shows the PSNR and
MSE values for test images using JPEG and Proposed Method.
Table-1a): Showing improved PSNR and MSE
Image Quality
JPEG Proposed Algorithm
PSNR MSE PSNR MSE
Barbara
Q=10 32.1245 39.8685 32.1252 39.8624
Q=5 30.8581 53.3670 31.3020 48.1817
Q=1 29.8702 66.9973 30.3731 59.6722
Peppers
Q=10 34.6592 22.2414 35.1713 19.7675
Q=5 32.1886 39.2840 33.1044 31.8158
Q=1 30.8311 53.6992 31.8167 42.7967
Baboon
Q=10 30.3658 59.7729 30.3536 59.9406
Q=5 29.6766 70.0523 29.9106 66.3775
Q=1 29.3194 76.0573 29.5498 72.1272
Elaine
Q=10 34.5543 22.7849 35.2924 19.2239
Q=5 32.3426 37.9154 33.4897 29.1146
Q=1 30.7786 54.3522 31.7557 43.4025
Table-1b): Showing improved PSNR and MSE
Image Quality
JPEG Proposed Algorithm
PSNR MSE PSNR MSE
Bridge
Q=10 30.9221 52.5855 30.9410 52.3582
Q=5 30-0113 64.8560 30.2752 61.0319
Q=1 29.2963 74.4631 29.5761 71.6927
Pentagon Q=10 31.9616 41.3920 32.2033 39.1513

33
Q=5 30.8437 53.5437 31.3752 47.3764
Q=1 29.7467 68.9296 30.1600 62.6725
Lena
Q=10 35.4089 18.7149 35.7721 17.2135
Q=5 32.6314 35.4763 33.5672 28.5993
Q=1 30.9227 52.5793 31.7384 43.5749
It is clear from the above table that the algorithm increases the PSNR and decreases the MSE
value for quality parameter (Q =10, 5, 1).
5. CONCLUSION AND FUTURE SCOPE
It is very much clear from the displayed results that the artifacts are removed to some extent as it
has increased the subjective as well as objective quality of the images. The algorithm effectively
reduces the visibility of blocking artifacts along with the preservation of edges. It also increases
the PSNR value of the image. As shown, the blocking artifacts are not removed totally. It is
because of the fact that the information lost in the quantization step is irrecoverable. The
algorithm only deals with pixel values (spatial domain) and the algorithm only tries to manipulate
the pixel values on the basis of some criteria.
The extent of blocking artifacts can also be reduced by manipulating the DCT coefficients, as
quantization is applied on the DCT coefficients. But frequency domain is having higher
complexity and consumption of time. There is always a tradeoff between time (complexity) and
efficiency (quality). Spatial domain is chosen where time (complexity) is the main concern, and
on the other hand frequency domain is preferred where efficiency is given more value. The extent
of reduction of blocking artifacts can be increased by recovering the information loss by using
some sort of prediction algorithm. It can be done by some learning technique (artificial
intelligence) or fuzzy logic. Further steps that can be taken in future are –
• The range of bit- rates can be extended in future.
• The postprocessor can be made generic postprocessor.
• It is possible to extend the proposed algorithm to video coding.
• Our proposed technique is restricted only to gray scale images, this can be extended
to color images.
• Extension towards real-time compression by designing faster heuristic for estimating
the interlayer dependencies.
• Design of a metric for measuring the blocking artifacts.
REFERENCES
[1] AHUMADA A. J.(1995) “Smoothing DCT Compression Artifacts” SID Digest, Vol. 25, pp. 708-711,
1995.
[2] ARIA Nosratinia (2002) Department of Electrical Engineering, University of Texas at Dallas,
“Enhancement of JPEG-Compressed Images by Re-application of JPEG” ©2002 Kluwer Academic
Publishers. Printed in the Netherlands
[3] AVERBUCH Amir Z., SCHCLAR Alon and DONOHO David L. (2005) “Deblocking of Block-
Transform Compressed Images Using Weighted Sums of Symmetrically Aligned Pixels” IEEE
Transactions on Image Processing, Vol.14, No.2, February 2005.

34
[4] CHANG Jyh-Yenng and LU Shih-Mao ,“Image Blocking Artifact Suppression by the Modified
Fuzzy Rule Based Filter” National Chiao Tung university, Hsinchu, Taiwan, ROC
[5] GARG Ankush and RASIWASIA Nikhel ,Term Paper“Reducing Blocking Artifacts in Compressed
Images” under Dr.Sumana Gupta,IIT Kanpur, pp. 6-10,15-16.
[6] GONZALEZ Rafeal C. , WOODS Richard E , “Digital Image Processing”, Second Edition, Pearson
Publication.
[7] GOTHUNDARUMUN A., WHITUKER RT and GREGOR J. (2001) “ Total Variation for the
Removal of Blocking Effects in DCT Based Encoding”, 0-7803-6725- 1/01/$10.00 02001 IEEE
[8] HYUK Choi and TAEJEONG Kim(2000) “Blocking-Artifact Reduction in Block-Coded Images
Using Wavelet-Based Subband Decomposition”, IEEE Transactions on Circuits and Systems for
video technology, Vol. 10, No 5,August 2000.
[9] ISMAEIL Ismaeil R. and WARD Rabab K. (2003) “Removal of DCT blocking Artifacts Using DC
and AC Filtering” 0-7803-797tL0/03/$l7.00 © 2003 IEEE, pp.229-232.
[10] KIRYUNG Lee, DONG Sik Kimand TAEJEONG Kim (2005) “Regression-Based Prediction for
Blocking Artifact Reduction in JPEG-Compressed Images” IEEE Transactions on Image Processing,
Vol.14, No.1, January 2005,pp-36-48.
[11] KIZEWSKI Lukasz (2004) “Image Deblocking Using Local Segmentation” Paper for Bachelor of
Software Engineering with Honours (2770) under Dr. Peter Tischer, Monash University ,November,
2004, kizewski@csse.monash.edu.au
[12] KWON Kee-Koo , LEE Suk-Hwan ,LEE Jong- Won, BAN Seong- Won , PARK Kyung-Nam , and
LEE Kuhn-I1 (2001) “Blocking Artifacts Reduction Algorithm in Block Boundary area using Neural
Network”,0-7803-7090-2/0 1 /$ 10.00 © 2001 IEEE
[13] LIEW Alan W.-C. and YAN Hong (2004) “Blocking Artifacts Suppression in Block-Coded Images
Using Overcomplete Wavelet Representation”, IEEE Transactions on Circuits and Systems for video
technology, Vol 14,No 4,April 2004
[14] LUO Ying and WARD Rabab K. “Removing the Blocking Artifacts of Block-Based DCT
Compressed Images”
[15] NALLAPERUMAL Krishnan, RANJANI Jennifer J., CHRISTOPHER Seldev (2006) , “Removal of
blocking artifacts in JPEG compressed images using Dual Tree Complex Wavelet Filters for
Multimedia on the Web”, 1-4244-0340-5/06/$20©2006IEEE
[16] NIE Yao, KONG Hao-Song ,VETRO Anthony and BARNER Kenneth (2005) “Fast Adaptive Fuzzy
Post-Filtering for Coding Artifacts Removal in Interlaced Video” Mitsubishi Electric Research
Laboratories, December 2005.
[17] PARK Kyung-Nam , KWON Kee-Koo , BAN Seong- Won and LEE Kuhn-I1 (2001) “Blocking
Artifacts Reduction in Block-Coded Images Using Self-similarity”,0-7803-7090-2/01/$10.00 0 2001
IEEE,pp.1667-1670.
[18] ROBERTSON Mark A. and STEVONSON Robert L. “DCT Quantization Noise in Compressed
Images”
[19] TUAN Q., Pham Lucas J., VAN Vliet “Blocking artifacts removal by a hybrid filter method”
[20] TRIANTAFYLLIDIS G.A., SAMPSON D., TZOVARAS D. and STRINTZIS M.G. , “Blockiness
Reduction in JPEG Coded Images”,0-7803-7503-3/02/$17.00©2002 IEEE,pp.1325-1328.

35
[21] TRIANTAFYLLIDIS G.A., TZOVARAS D. and STRINTZIS M.G. (2001) “Detection of Blocking
Artifacts of Compressed Still Images” 0-7695-1183-x/01 $10.00©2001 IEEE,pp.607-610.
[22] TRIANTAFYLLIDIS George A., TZOVARAS Dimitrios , SAMPSON Demetrios, Strintzis Michael
G. (2002) “Combined Frequency and Spatial Domain Algorithm for the Removal of Blocking
Artifacts” EURASIP Journal on Applied Signal Processing 2002:6, 601–612
[23] WANG Zhou, BOVIK Alan C., and EVANS Brian L. , “ Blind Measurement of Blocking Artifacts in
Images”
[24] WANG Ci,, ZHANG Wen-Jun, and FANG Xiang-Zhong (2004), “Adaptive Reduction of Blocking
Artifacts in DCT Domain for Highly Compressed Images”, 0098 3063/04/$20.00 © 2004 IEEE
[25] WANG Ci, XUE Ping , Weisi LIN, ZHANG Wenjun , and YU Songyu (2005) “Fast Edge-Preserved
Postprocessing for Compressed Images” IEEE Transactions on Circuits and Systems for Video
Technology, Vol. 16, No. 9, September 2006,pp. 1142-1147.
[26] WENFENG Gao, COSKUN Mermer, and YONGMIN Kim(2002) “A De-Blocking Algorithm and a
Blockiness Metric for Highly Compressed Images” IEEE Transactions on Circuits and Systems for
Video Technology, Vol. 12, No. 12, December 2002,pp.1150-1159.
[27] YANG Fu-zheng, WAN Shuai, CHANG Yi-lin, LUO Zhong (2006) “A no-reference blocking artifact
metric for B-DCT video” Journal of Zhejiang University Received Nov. 30, 2005; revision accepted
Feb. 17, 2006,ISSN 1009-3095 (Print); ISSN 1862-1775 (Online).pp.95-100.
[28] YANG Yongyi, GALATSANAS Nikolas P. and KATSAGGELAS Aggelas K., (1995) “Projection-
Based Spatially Adaptive Reconstruction of Block Transform Compressed Images”,IEEE Transaction
on Image Processing, Vol4 as 7th July 1995,pp.421-431.
Author
Sukhpal Singh obtained his B.Tech. (Computer Science and Engineering) Degree
from G.N.D.E.C. Ludhiana (Punjab) in 2010. He joined the Department of
Computer Sci. Eng. at North West Institute of Engineering and Technology,
Moga (Punjab) in 2010. Presently he is pursuing M.E. (Software Engineering)
degree from Thapar University, Patiala. His research interests include Image
Compression, Software Engineering, Cloud Computing, Operating System and
Database.

An Algorithm for Improving the Quality of Compacted JPEG Image by Minimizes the Blocking Artifacts

Recommended

More Related Content

What's hot (20)

Similar to An Algorithm for Improving the Quality of Compacted JPEG Image by Minimizes the Blocking Artifacts (20)

Recently uploaded (20)

An Algorithm for Improving the Quality of Compacted JPEG Image by Minimizes the Blocking Artifacts