International Journal of Computer Applications (0975 – 8887)

Volume 42– No.6, March 2012

Blocking Artifacts Reduction in Block based Discrete

Cosine Transform Compressed Images

Parmjeet Kaur Poonam Sethi

Assistant Professor, Assistant Professor,
Department of Electronics and Department of Electronics and
Communication Engineering, Communication Engineering,
College of Engineering and DAV Institute of Engineering
Management, and Technology,
Kapurthala. Jalandhar

ABSTRACT introduce compression artifacts and are especially suitable for

Image compression is a technique to reduce the volume of natural images such as photographs in applications where
information to be transmitted. Compression technique reduces minor loss of fidelity is acceptable to achieve a substantial
the bandwidth and the volume of the information to be reduction in bit rate. With the development of signal
transmitted. We can compress audio signal, video signal, text, processing, compression technology has been widely applied
fax and images. For medical images lossless compression is in various applications, including digital cameras, DVD and
used and for other types lossy compression can be used. For broadcast. The process of image compression is required in
compressing an image we can use the DCT technique. But many imaging and video applications. However, these
after compression during decompression and recovering applications are all bothered by an annoying problem i.e.
original image from compressed image we can face problem blocking artifacts.
of blocking artifacts. Various methods can be used for
removing blocking artifacts. One of them is DCT filtering. 2. BLOCKING ARTIFACTS
Further for getting better results we can remove blocking Blocking artifacts are the major coding artifacts caused by
artifacts by spatial and hybrid filtering method. Our high compression. These appear as grid noise along the block
experimental results shows that hybrid filtering gives better boundaries in smooth areas and are caused due to independent
performance on the bases of better PSNR, BER and MSE. encoding of each block without considering the correlation
between adjacent blocks. To reduce irrelevancy, redundancy
Keywords of the image data and volume of data to be transmitted, image
DCT, Spatial filtering, Hybrid filtering, JPEG, Blocking compression is must. However with the use of lossy image
artifacts, image compression, MSE, BER, PSNR. compression methods blocking artifacts get introduced. A
compression artifact is a noticeable distortion of media which
1. INTRODUCTION may be an image, audio or video, due to the application of an
Image processing is the method to improve the original overly aggressive or inappropriate lossy data compression
images obtained from camera or any other source. Many algorithm. These lossy data compression schemes discard
image processing techniques can be used. One of the some data to simplify the media sufficiently to store it in the
processing techniques, image compression is used to reduce desired space (data-rate). If there is not enough data in the
the amount of data required to represent an image. compressed version to reproduce the original with acceptable
Compressing an image is significantly different than fidelity, artifacts will result. Alternatively, the compression
compressing raw binary data. This is because images have algorithm may incorrectly determine certain distortions to be
certain statistical properties which can be exploited by of little subjective importance but they may in fact be
encoders specifically designed for them. Also, some of the objectionable to the viewer. The artifacts such as over
finer details in the image can be ignored for the sake of saving smoothing of images, degradation due to quantization errors,
a little more bandwidth or storage space. The objective of ringing effects, blurring of images, blocking noise, image
image compression is to reduce irrelevance and redundancy of features irregularities etc. So to overcome these artifacts,
the image data in order to be able to store or transmit (text, various methods are used such as Normal DCT filtering,
fax, images) data in an efficient form. Now a day’s Transform Spatial Filtering, Hybrid Filtering methods, DWT, Adaptive
coding is the most commonly used method for this. In Fuzzy Filtering method, iterative reconstruction algorithm etc.
transform coding, Fourier related transform such as DCT or The boundary regions between blocks are identified as either
wavelet transform are applied, followed by quantization and smooth or non-smooth regions [2]. The blocking artifacts in
entropy coding. Compression artifacts occur in many common smooth regions are removed by modifying a few DCT
media such as DVD’s, common computer file formats such as coefficients appropriately, whilst an edge-preserving
JPEG, MP3 and MPEG files. Image compression may be smoothing filter is applied to the non-smooth regions. Above
lossy or lossless [1]. Lossless compression is preferred for mentioned algorithms are highly efficient in removing
archival purposes and often for medical imaging. Lossy blocking artifacts and produce visuality that is much more
compression methods, especially when used at low bitrates, pleasing to the eye.

19
 International Journal of Computer Applications (0975 – 8887)
Volume 42– No.6, March 2012

3. METHODS FOR REMOVAL OF 3.4 Application of DCT in Loss Image

BLOCKINGARTIFACTS Compression
The DCT is used for two international image and video The DCT transform coding method compresses image data by
compression standards, Joint Photographic Experts Group representing the original signal with a small number of
(JPEG) and Motion Picture Experts Group (MPEG).This transform coefficients. It exploits the fact that for typical
method are computationally faster and can efficiently images a large amount of signal energy is concentrated in a
eliminate blocking effects for images coded at different bit small number of coefficients. The goal of DCT transform
rates without producing noticeable artifacts. It smoothes out coding is to minimize the number of retained transform
the undesired block edges while retaining the sharpness of the coefficients while keeping distortion at an acceptable level.
decoded image to a noticeable degree. Its algorithm requires a This DCT is done in 8x8 non overlapping blocks. Block
fairly low computational effort [9-10]. Thus it is useful for transform coding divides an image into blocks of equal size
real-time applications. and processes each block independently. Block processing
3.1 Discrete Cosine Transform allows the coder to adapt to local image statistics, exploit the
correlation present among neighboring image pixels and to
reduce computational and storage requirements. Larger blocks
DCT is a Fourier-related transform similar to the discrete do not offer significantly better compression, since correlation
Fourier transform (DFT), but using only real numbers. DCTs between pixels tends to decrease as the block size grows. To
are equivalent to DFTs of roughly twice the length, operating get best image quality at a given bit- rate (or compression
on real data with even symmetry, while in some variants the rate) is the main goal and this can be achieved by using
input and output data are shifted by half a sample. The DCT is different DCT functions for image compression and for
used in JPEG image compression, and MPEG. There, the two- different blocking artifacts reduction methods.
dimensional DCT-II of [N × N] blocks is computed and the
results are quantized and entropy coded. In this case, N is 4. LITERATURE STUDY
typically 8 and the DCT-II formula is applied to each row and There have numerous numbers of researches on image
column of the block. The result is an [8 × 8] transform compression and blocking artifacts reductions. There are
coefficient array.DCT is very efficient method used in image number of filtering techniques. An advanced discrete cosine
compression and processing of images .It is easy to implement transform (DCT)-based image compression method that
this method in MATLAB. Its mathematical equation is easy to combined advantages of several approaches. Image
implement through MATLAB programming. compression using Discrete Cosine Transform (DCT) is one
of the simplest commonly used compression methods. The
3.2 Spatial filtering quality of compressed images is marginally reduced at higher
compression ratios due to the lossy nature of DCT
Spatial filtering is a term used to describe the methods used to
compression and an optimum DCT compression ratio. And an
compute spatial density estimates for events observed at
ideal image compression system that yield high quality
individual locations. Spatial prediction does not out-perform
compressed images with good compression ratio, while
pure DCT based technique (such as JPEG) in terms of bit-rate
maintaining minimum time cost. Firstly divide an image into
tradeoff. However, at very low bit rates it results in far fewer
blocks of different sizes by a rate-distortion-based modified
blocky artifacts and markedly better visual quality. It
horizontal-vertical partition scheme. Statistical redundancy of
describes a set of tools for displaying functions estimated
quantized DCT coefficients of each image block is reduced by
from these data points that are distributed in two-dimensional
a bit-plane dynamical arithmetical coding with a sophisticated
space. The algorithm employs a two-dimensional (2-D) filter
context modeling. This post-filtering removes blocking
in the areas away from edges, and for near edges, one-
artifacts in decompressed images. This method provides
dimensional (1-D) filter aligned parallel to edge so as to
significantly better compression than JPEG and other DCT-
reduce the blocking artifacts. So for removing blocking
based techniques. Image compression is a very important
artifacts near edges we need spatial filtering.
issue for several applications in the area of multimedia
communications, the objective being reduction of storage and
3.3 Hybrid filtering transmission costs. However, the reconstructed images from
Hybrid filtering method is used to combine both DCT and JPEG compression produce annoying blocking artifacts near
spatial filtering methods discussed already in previous section. block boundaries, particularly in highly compressed images.
There is no need of IDCT in hybrid filtering method. So The blocking artifacts in smooth regions were removed by
hardware implementation of this method is easy. Ringing modifying a few DCT coefficients appropriately, whilst an
effect can be reduced by hybrid filtering method. This method edge-preserving smoothing filter was applied to the non-
can give better performance in terms of both objective and smooth regions. Block edge artifacts is one of the most
subjective views than the previous methods. This method can noticeable impairment associated with MPEG and other block
be seen as a hybrid approach, because it is applied to the based encoding techniques. Blocking artifacts can be removed
blocks with vertical or horizontal edge [14]. The hybrid using variable width non-linear filtering as well as
method works by combining the filtering methods for three randomized masking of geometric patterns associated with
different grayscale gradient ranges respectively. It is used to blockiness. High-frequency details of the coded images are
get better processing speed, reduce blurring and enhance mainly contaminated by quantization noise. Non-linear
edges to a large extent. Therefore it can be used for high- adaptive filter to reduce both staircase and grid noises in
speed noise elimination of 3D images with large data. We can decompressed images without any additional information and
implement this method by MATLAB. without introducing significant blurring. The reconstructed
images from JPEG compression produced noticeable image
degradation near the block boundaries, in particular for highly
compressed images, because each block was transformed and

20
 International Journal of Computer Applications (0975 – 8887)
Volume 42– No.6, March 2012

quantized independently. It was an adaptive approach which

performed blockiness reduction in both the DCT and spatial
domains to reduce the block-to-block discontinuities. This
post- processing algorithm, which consisted of three stages,
reduced these blocking artifacts efficiently.
5. PROBLEM FORMULATION
The image compression is used to minimize the number of
bits needed to represent an image without sacrificing quality
of image. However, when we compress the information, then
we face the problem of unwanted blocking artifacts. Blocking
artifacts can take on several forms in DCT compressed
images. These artifacts are visually very annoying and have a
substantial impaction to the subjectively perceived image
quality. The artifacts removal algorithms are cost effective
and yield better performance in terms of both objective and
subjectively. Fig 1: Original image before compressed.
5.1 Blocking Artifacts reduction algorithms
Blocking Artifacts reduction algorithms give significant Original Image Compression Factor 2

improvement in the separation between blocking noise, image

features and effective reduction of image blurring. Many steps 50 50
are to be performed to preserve the global and local edges of
the processed image, remove blocking noise and ensure 100 100
smoothness without blurring. It helps to remove blocking
artifacts and to enhance feature regularities [18]. In general 150 150
the DCT based method required as much as six times more
processing time as compared to the spatial method. The time 200 200
in DCT methods will be always more as a lot of processing is
50 100 150 200 250 300 350 400 50 100 150 200 250 300 350 400
involved in conversion to and from the DCT coefficients.
Spatial filtering is an effective method for improving JPEG,
MPEG compression in fixed and variable rate control Compression Factor 4 Compression Factor 8
applications. The method Hybrid Filtering yields better
performance in terms of both objective and subjective views
50 50
than the conventional methods. Hybrid filtering method is
nonlinear, computationally efficient, and spatially adaptive.
100 100
5.2 Steps for the Research Work
Step1. To compress image using DCT technique. 150 150

Step2.To remove blocking artifacts using DCT filtering,

spatial filtering and hybrid filtering method. 200 200

Step3.To compares the above techniques qualitatively and

50 100 150 200 250 300 350 400 50 100 150 200 250 300 350 400
quantitatively on the basis of parameters such as PSNR, MSE
and BER.
Table 1. Various Parameters
Fig 2: Images obtained after DCT compression, with
compression factor 2, 4, & 8 respectively.
Filtering MSE PSNR BER
Technique.

DCT 30.5695 33.2779 0.0300

filtering

Spatial 11.4668 37.5364 0.0266

Filtering

Hybrid 0.8204 48.9905 0.0204

filtering

Fig 3: Image having Blocking Artifacts.

21
 International Journal of Computer Applications (0975 – 8887)
Volume 42– No.6, March 2012

Table 2. Table showing Images recovered after Filtering by three different methods and their graphs of MSE, BER&PSNR.

Image Recovered Graph of various parameters.

mse-ber-psnr
35

30

25

20

15

10

0
1 2 3

Fig4: MSE, BER &PSNR of DCT filtering.

Fig 3 :Image recovered by DCT Filtering

mse-ber-psnr
40

35

30

25

20

15

10

0
1 2 3

Fig 5: Image recovered by Spatial Filtering

Fig 6: MSE, BER &PSNR of Spatial filtering.

mse-ber-psnr
50

45

40

35

30

25

20

15

10

0
1 2 3

Fig 7: Image recovered by Hybrid filtering. Fig 8: MSE, BER &PSNR of Hybrid filtering.

22
 International Journal of Computer Applications (0975 – 8887)
Volume 42– No.6, March 2012

Table 3. Graphs showing comparison of various methods of Removing Blocking Artifacts.

Comparison of the two methods Comparison of the three methods

peak signal to noise ratio peak signal to noise ratio

40 50

45
35

40

30
35

25
30

25
20

20
15

15

10
10

5 5

0
0 1 2 3
1 2

Fig 9:Graph of PSNR of first two methods Fig 10:Graph of PSNR of three methods

mean square error mean square error

35 35

30 30

25 25

20 20

15 15

10 10

5 5

0 0
1 2 1 2 3

Fig 11:Graph of MSE of first two methods Fig 12:Graph of MSE of three methods

bit error rate bit error rate

0.035 0.035

0.03 0.03

0.025
0.025

0.02
0.02

0.015
0.015

0.01
0.01

0.005
0.005

0
1 2
0
1 2 3

Fig 13: Graph of BER of first two methods

Fig 14: Graph of BER of three methods

23
 International Journal of Computer Applications (0975 – 8887)
Volume 42– No.6, March 2012

6. EXPERIMENTAL RESULTS JPEG Compressed Images”, IEEE Transactions on

Image Processing, Vol. 10, No. 9, pp. 1057–7149, Sep.
By using MATLAB R2010a version 7.10.0 software we 2001.
observed the results, that are shown in Table1, Table2 and
Table3 respectively. Firstly we take the original image then [5] Massinio Mancuso, Monica Besana, RinaldoPoluzzi, “A
compress it using DCT compression we concluded that as we new post processing algorithm To reduce block artifacts
increase the compression factor the blurring of the image in block coded images”, IEEE Microelectronics, Vol. 43,
increases. After that when image is decompressed for pp.303-307, Aug. 1997.
recovering the original image, it shows blocking artifacts in it. [6] Shizhong Liu, Student Member, and Alan C. Bovik,
To remove those blocking artifacts we need to use the filtering Fellow,”Efficient DCT-Domain Blind Measurement and
techniques. So we use the filtering technique DCT, Spatial Reduction of Blocking Artifacts”, IEEE Circuits &
and hybrid.Table1, shows the various parameters like BER Systems for video Technology, Vol.12, No.12, pp.1139-
(Bit Error Rate), MSE (Mean Square Error), PSNR (Peak 1149, Dec. 2002.
Signal to Noise Ratio) of these three techniques.Table2,
shows the output of three techniques and graphical [7] F. M. Bayer and R. J. Cintra Member, “Image
representation of various parameters.Table3, shows the Compression via a Fast DCT Approximation”, IEEE
graphical representation of the comparison of the BER, Latin America Transactions, Vol. 8, No. 6, pp.708-713,
PSNR, MSE .Where PSNR is peak signal to noise ratio in Dec. 2010.
decibels, between the two images. This ratio is often used as a [8] Ephraim Feigand Heidi Peterson IBM Research
quality measurement between the original image and the Yorktown Heights, NY, “Image Compression using
compressed image. The higher the PSNR, the better the spatial prediction”,ICASSP, Vol. 4, pp.2339-2342, May
quality of the compressed or reconstructed image. MSE& 1995.
PSNR are the two error matrices used to compare image
compression quality. The MSE represents the cumulative [9] Ee-Leng Tan1, Woon-Seng Gan1 and Meng-Tong
squared error between the compressed and original image, Wong, “Fast Image Resizing in Discrete Cosine
whereas PSNR represents the measure of the peak error. The Transform Domain with Spatial Relationship between
lower the value of MSE the lower the error.BER is bit error DCT Block and Its Sub-Blocks”, IEEE ICALIP, Vol.1,
rate and we determine how many of the received bits are in pp. 1683-1687, July 2008.
error, and then compute the bit-error-rate as the number of bit
[10] G. A. Triantafyllidis, D. Tzovaras, D. Sampson, and
errors divided by the total number of bits in the transmitted
signal. M.G.Strintzis, “A hybrid algorithm for, the removal of
blocking artifacts”, IEEE Multimedia & Expo. , Vol.1,
6.1CONCLUSION pp.161-164, Nov. 2002.

We concluded that both qualitatively and quantitatively the [11] Sang-Jun Park and JechangJeong, Member, “Hybrid
spatial filtering technique is better than the DCT filtering and Image Up sampling Method in the Discrete Cosine
hybrid filtering is better than the spatial filtering. Because Transform Domain”,IEEE Transactions on Consumer
when we compare first two techniques then we observed that Electronics, Vol.56, No. 4, pp.2615-2622, Nov. 2010.
PSNR of second method is more than first and MSE and BER [12] Wang Rui, Lin Jiangli, Li Deyu, “Edge Enhancement
are less. And when we compare the three methods than we and Filtering of Medical Ultrasonic Images using a
observed that hybrid is having good PSNR and lesser value of Hybrid Method”, ICBBE Bioinformatics and Biomedical
BER and MSE than the first two methods. As we know that Engineering, Vol.1, pp.876 – 879, July 2007.
for better filtering and recovering of images we need more
PSNR and less BER and MSE so hybrid filtering is better than [13] I. O. Kirenko, R. Muijs, L. Shao,“New method for the
the both methods. reduction of artifacts in MPEG compressed video
sequences”, IEEE Multimedia & Expo.,Vol.7,No.1,
7. REFERENCES pp.469-472, Dec.2006.
[1] Nikolay N. Ponomarenko, Karen O. Egiazarian, Vladimir
[14] Shuanhu Wu, Hong Yan, and Zheng Tan,“An Efficient
V.Lukin, “High-Quality DCT-Based Image Compression
Wavelet-Based Deblocking Algorithm for Highly
Using Partition Schemes”, IEEE Signal Processing
Compressed Images”, IEEE Transactions on Circuits &
Letters, Vol. 14, No. 2, pp.105-108, Feb. 2007.
Systems for Video Technology.,Vol. 11, No.11, pp.
[2] Sukhwinder Singh, Vinod Kumar, H.K. Verma, 1193-1198, Nov.2001.
“Reduction of blocking artifacts in JPEG compressed
[15] Ying Luo and Rabab K. Ward, Fellow, “Removing the
images”, Digital Signal Processing, Vol. 17, pp. 225–
Blocking Artifacts of Block-Based DCT Compressed
243, Jan. 2007.
Images”, IEEE Transactions on ImageProcessing, Vol.
[3] Ekaterina BarzykinaRabab K. Ward ,“Removal of 12, No. 7, pp. 838-842, July 2003.
Blocking Artifacts using Random Filtering”,
[16] Kawaldeep Singh,Parveen Kumar, “Algorithm for
IEEEimage processing, Vol. 2, pp. 904-908,
Blocking Artifact Detection & Reduction using adaptive
oct.1999.
filtering in Compressed Images”, IJAEST, Vol. 5, No. 2,
[4] Amjed S. Al-Fahoum and Ali M. Reza, “Combined Edge pp.156-162, 2010.
Crispiness and Statistical Differencing for Deblocking

24