International Journal of Image and Graphics

Vol. 20, No. 2 (2020) 2050007 (16 pages)

#.c World Scienti¯c Publishing Company
DOI: 10.1142/S0219467820500072
by UPPSALA UNIVERSITY on 05/18/20. Re-use and distribution is strictly not permitted, except for Open Access articles.

The PHY-NGSC-Based ORT Run Length Encoding

Scheme for Video Compression

Poorva Girishwaingankar*
Electronics Department
Int. J. Image Grap. 2020.20. Downloaded from www.worldscientific.com

Thakur College of Engineering and Technology

Mumbai, India
[email protected]

Sangeeta Milind Joshi

Electronics Department
Vidyalankar Institute of Technology
Mumbai, India

Received 18 January 2019

Accepted 30 August 2019
Published 24 April 2020

This paper proposes a compression algorithm using octonary repetition tree (ORT) based on
run length encoding (RLE). Generally, RLE is one type of lossless data compression method
which has duplication problem as a major issue due to the usage of code word or °ag. Hence,
ORT is o®ered instead of using a °ag or code word to overcome this issue. This method gives
better performance by means of compression ratio, i.e. 99.75%. But, the functioning of ORT is
not good in terms of compression speed. For that reason, physical- next generation secure
computing (PHY-NGSC) is hybridized with ORT to raise the compression speed. It uses an
MPI-open MP programming paradigm on ORT to improve the compression speed of encoder.
The planned work achieves multiple levels of parallelism within an image such as MPI and open
MP for parallelism across a group of pictures level and slice level, respectively. At the same time,
wide range of data compression like multimedia, executive ¯les and documents are possible in
the proposed method. The performance of the proposed work is compared with other methods
like accordian RLE, context adaptive variable length coding (CAVLC) and context-based
arithmetic coding (CBAC) through the implementation in Matlab working platform.

Keywords: Video compression; discrete cosine transform; quantization; run length encoding;
octonary repetition tree; hybrid MPI-open MP strategy.

1. Introduction
The redundant information is converted into single data to make lossless data
compression. The graphic interchange format (GIF) is used as the replacement of
black and white image compression. This procedure is arranged with the support of
CompuServe method to RLE based early graphics ¯le format.1 CompuServe is one
* Corresponding author.

2050007-1
 P. Girishwaingankar & S. M. Joshi

of the online services which is successor of GIF and can also be considered as a
standard image compression technique.2 The transformation and quantization of
image blocks is the second step of JPEG based continuous tone images in all kind
of photographs.3 The redundancy of the information present in the video can be
eliminated by video compression methods.
by UPPSALA UNIVERSITY on 05/18/20. Re-use and distribution is strictly not permitted, except for Open Access articles.

Temporal motion compression and spatial image compression are two main
techniques jointly used in video compression algorithms.4 The lossy compression
plays major role in video compression techniques. Compression, decompression, cost,
video quality and system needs are some of the trade-o® in lossy compression tech-
niques.5,6 Continuous connection of still image frames creates video data.7 The pixel
variation mainly based on the encoding and variation related workings.8 Macro
Int. J. Image Grap. 2020.20. Downloaded from www.worldscientific.com

blocks are basically formed from the square-shaped neighboring pixels in certain kind
of video compression schemes.9,10
The encoding of similarities between the large frames can be done by storing
attitude of human visual perception.11 It is very di±cult to identify the brightness
changes than the color di®erences in an image or video.12 The space reduction of
compression algorithms is totally same as that of JPEG image compression.13–16
Several techniques are available for video compression and one of them is moving
picture experts group (MPEG) which is ISO/ITU standard for digital video
compression.
Pixels of input image may be permanently neglected while lossy compression
procedure for MPEG technique.17–19 There are several applications available which
are implemented based on JPEG methods along with the elimination of redundant
data.20 Even in the presence of various techniques for video compression, most of
them not exhibit good performance by means of compression ratio. So proposed work
concentrate on development of new technique for video compression standards.
The overall methodology for planned research is organized as follows, Sec. 2
contains some of the working related to the proposed work. The basic steps of video
compression are organized in Sec. 3. Section 4 contains the detail description of
proposed ORT based hybrid MP-open MPI methodology for video compression. The
related performances are compared with the existing methods and clearly explained
in Sec. 5. Finally, the whole process is concluded in the conclusion part.

2. Related Works
Harisha and Murugendrappa21 had proposed CAVLC scheme for H.264 encoder.
Transform coe±cient coding had been performed in CAVLC method. Code selection
process follows context adaption on the basis of newly coded coe±cients. Scan or-
dered and residual encoded blocks are utilized in CAVLC. Quantization processor
and integer transformation are responsible for generation of quantized coe±cient
inputs. A simulation result shows good compression performance with high picture
resolution.

2050007-2
 The PHY-NGSC-Based ORT Run Length Encoding Scheme for Video Compression

Yang et al.22 used CBAC technique to achieve lossless compression. CBAC aided
RLE provides admirable coding e±ciency without using unnecessary contexts.
CBAC makes RLE as compatible and accepted as normative lossless compression by
MPEG group. For MPEG-G applications, CBAC combined with context adaptive
binary arithmetic coding to give better performance.
by UPPSALA UNIVERSITY on 05/18/20. Re-use and distribution is strictly not permitted, except for Open Access articles.

Temporal Fluctuation Reduced Video Encoding (TFRE) has been proposed by

Kong and Dai et al.23 which focus to enhance performance of self-video analysis
process in surveillance system. Sum of absolute frame di®erence (SFD) integration
enables TFRE to suppress °uctuations in compressed videos. Vastly accurate object
identi¯cation with lower bit rate and high quality is result of TFRE.
Ahmad and Siddiqui24 developed compression algorithm on the basis of accordian
Int. J. Image Grap. 2020.20. Downloaded from www.worldscientific.com

RLE. Complexity reductions have also been achieved along with competence im-
provement. Use of accordian algorithm modi¯es RLE process in video surveillance
compression. New method of loop ¯lter for video compression on the basis of soft
threshold noise compression has been proposed in Zhang et al.25 However, every
existing methods exhibit extreme computational complexity. Therefore, physical-
next generation secure computing (PHY-NGSC)-based ORT-RLE scheme developed
in proposed work to achieve better performance with reduced computation di±culty.

3. Basics of Video Compression

The process of reducing the amount of information from an image without a®ecting
the quality of the image. The size reduction can be achieved through the utilization
of suitable compression methods with little or no adverse e®ect on visual quality.
There series of frames presented in the video ¯le and hence image compression
techniques are applied on each frame. The several techniques used for the com-
pression purpose are listed as follows,

. Block transform
. Quantization
. Coding
. YUV and RGB color spaces
. Motion estimation and compensation

Generally, the ¯rst step of the image compression is the division of an image into
several blocks (8
8 of pixels). Next, unwanted information can be extracted
through the quantization process. Then the DCT coe±cient is encoded and most of
the DCT coe±cients are zero at high frequency. The group of zeros is encoded in a
zigzag fashion in the process of run length encoding. Generally, this RLE method
contains the duplication issue due to the usage of code word and °ag. Hence, RLE
with Octonary repetition tree (ORT) algorithm is proposed in this research work for
the sake of overcoming this problem. As well as a hybrid MPI-Open MP strategy is
suggested along with ORT method to enhance the compression speed. Normally, the
input video signal should be sampled before giving it to the RGB to YUV converter.

2050007-3
 P. Girishwaingankar & S. M. Joshi

Buffer Fullness

Residual ORT based RLE with

RGB to YUV - DCT Quantization hybrid MPI-Open MP Buffer
strategy
by UPPSALA UNIVERSITY on 05/18/20. Re-use and distribution is strictly not permitted, except for Open Access articles.

Sub-Sampling MV data
Output
Inverse
Bitstream
Quantization

Input Video
Signal
Inverse DCT

MC- prediction
+
Int. J. Image Grap. 2020.20. Downloaded from www.worldscientific.com

Frame Store
Motion
Compensation Previous
Reconstructed
MV data Frame
Motion
Estimation

Fig. 1. Architecture of a video encoder.

The color places like RGB and YUV are used to indicate the color images.
The process of storing di®erence of the previous frame instead of taking entire frame
variation is the main function of motion compensation and estimation process.
The general architecture of normal video encoder is shown in Fig. 1. Here, the
input video signal kept under the sub sampling methods to remove the unwanted
noise. Then individual frames are given as an input to the color sampling methods.
The heavy correlation between the successive frames can be removed through the
application of motion estimation and compensation methods. These methods will be
applied to the results of both RGB to YUV and stored frames. The estimation and
transmission of motion vector (MV) at each block is responsible for the process of
motion compensation prediction (MC-prediction). It indicates the current position
of the image in the previously stored frames. The object mobility in an image
sequence can be examined through motion estimation process. This data can be used
by the motion compensator to perform the compression process.
The prediction error will be formed through the subtraction of motion compen-
sation prediction from the original image. It actually provides the di®erence between
the predicted macro block and the actual macro block being encoded. Then this
prediction error may be transformed with DCT for the sake of producing DCT
coe±cients. These coe±cients are further undergoing the quantization process to
reduce the number of bits for the coe±cient representation. The quantized DCT
coe±cients are given to ORT-based RLE coding with PHY-NGSC to reduce the
average number of bits per coe±cient. The internal loop contains the block of inverse

2050007-4
 The PHY-NGSC-Based ORT Run Length Encoding Scheme for Video Compression

quantization and IDCT. The predicted macro block from the frame store will be
added back on a residual in a pixel by pixel manner to serve as an MCP's reference
for predicting subsequent pictures. The data from the reference picture of the
memory can be matched with the decoder.
by UPPSALA UNIVERSITY on 05/18/20. Re-use and distribution is strictly not permitted, except for Open Access articles.

4. Proposed Methodology
Run length encoding (RLE) is a technique of replacing a set of similar data in the
compression algorithm. This algorithm does not require much CPU power due to its
simplicity for the implementation. But, this algorithm is very e®ective for certain ¯les
with plenty of repetitive information. For example, 5 9 6 1 2 is a sequence of di®erent
bytes and will be encoded as (5 1 9 1 6 1 1 1 2 1) by RLE which intern increases the
Int. J. Image Grap. 2020.20. Downloaded from www.worldscientific.com

data size and named as duplication problem. But, Octonary repetition tree (ORT)
based compression algorithm can be used as an alternative for reducing this kind of
issues. Lack of compression speed is the only limitation of ORT method. And hence, a
hybrid MPI-Open MP strategy or PHY-NGSC is used to speed up the compression
by dividing the data and pipelining the recursive runs.

4.1. Octonary repetition tree (ORT)-based RLE

Run length encoding is one type of lossless data compression method. The lossy
compression method is irreversible also leads to data missing based on the com-
pression level. But, the reconstruction of original data is possible in ORT based RLE
which come under the lossless image compression technique with duplication issue.
Therefore, ORT is proposed as an optimization algorithm which is added to
represent the sequence of points with the same values. It locates the sequence of the
same bytes with less overhead without using additional bytes to identify the number
of same bytes. There is no need of idle bytes in an image and no restriction for the
encoding process. Especially, the compression ratio is nearly about 100% for all type
of images such as wider data type, documents, executive ¯les and multimedia. In
ORT, the tree structure automatically produced while the data reviewing process.
Image Size ¼ Number of Pixels
Number of bytes per pixel: ð1Þ

Actually, one piece of information in a pixel can be named a bit. Commonly, a

pixel has more bits of results in higher resolution. For example, a pixel with 1, 8, 24
bits may contain 2 (black and white), 256 and 16 million possible colors, respectively.
The proposed algorithm mainly operates on the bits of pixels for reducing the du-
plication problem instead of using °ag or code word. Figure 2 shows the structure of
ORT. The ORT algorithm looks like a normal tree structure in which each node
represents one byte (8 bits) of a pixel.26 The expressions for measuring the parent
and child nodes of the tree is given as follows:
Parent ¼ ½ðI  1Þ=8 if I 6¼ 0
; ð2Þ
k th Child ¼ 8I þ k if 8I þ k  N

2050007-5
 P. Girishwaingankar & S. M. Joshi

b c d e f g h i

j k l
by UPPSALA UNIVERSITY on 05/18/20. Re-use and distribution is strictly not permitted, except for Open Access articles.

a is Root, b is the first child of a, c is the second

child of a, d is the third child of a, e is the fourth
child of a, f is the fifth child of a, g Is the sixth
child of a, h is the seventh child of a, i is the eight
child of a, j is the first child of b, k is the second
child of b, l is the third child of b…..

Fig. 2. Structure of ORT.

Int. J. Image Grap. 2020.20. Downloaded from www.worldscientific.com

where N is the total number of nodes in the tree and I indicates the index value
which is in the range of 0 to N  1 (0 to 255). k is the total amount of child node
present in the tree. Each parent node (byte) can be divided into 8 child (bits) motes
until reaches the leaves (node with no child) to denote the exact place of duplicated
bytes. The division of the tree will be stopped when the node has no children. File size
is responsible for the tree level calculation which can be measured based on total
amount of bytes (M ) in a ¯le.
Level of tree ¼ log8 ðMÞ: ð3Þ
Every bit in the leaf denotes the pairs of equality in a sequence. The zero bit value
says that the pair of bytes is dissimilar. The byte pair is the same for the bit value of 1
and hence one byte in the tree leaf displays the similarity of 8 bytes in data. The byte
comparison for each iteration of the algorithm with the total amount of iterations
ðKÞ is given as follows:
X
10 X
10
byte ½N ¼ byte ½N þ K þ 1: ð4Þ
K¼0 K¼0

In this way, the proposed algorithm eliminates the same bytes of the sequence
while passing the information and locating the places in the tree. This algorithm runs
for several times until it reaches the desired result by giving the output of one
iteration as an input for the next iteration for the encoding purpose. But the depth of
the tree is ¯xed and embedded for several applications and does not lead to further
process. The diagrammatical representation of ORT algorithm is given in Fig. 3.
Figure 3 shows the working °ow of the proposed algorithm in which the pixel size
(iÞ and iteration (kÞ are initialized as zero before the tree formation. In general, the
pixel values of each frame are stored in a matrix format with equal rows and columns.
Initially, the ¯rst row will come for the purpose of compression which can be assumed
as the variable of A. If the value of (iÞ is less than the pixel size, the tree will be
formed until it read the full sequence. Then the next sequence goes for the com-
pression by updating the ORT for every row of the pixel size. In this way, all frames

2050007-6
 The PHY-NGSC-Based ORT Run Length Encoding Scheme for Video Compression

Start

i=0
K=0
by UPPSALA UNIVERSITY on 05/18/20. Re-use and distribution is strictly not permitted, except for Open Access articles.

A=First row

Yes

Yes K<=iteration
i<=Pixel size ORT Formation
K++
Int. J. Image Grap. 2020.20. Downloaded from www.worldscientific.com

No
No No
No

No
No A [i]= A [i+K+1] Compressed file +
i++ ORT

Yes
Yes

Update ORT () Finish

Fig. 3. Flowchart for proposed algorithm.

in an image are compressed without using the °ag or code word to denote the
repeated data.
Uncompressed Size
Compression Ratio ðCRÞ ¼ : ð5Þ
Compressed Size
Equation (5) is used to calculate the compression ratio which is the ratio between
uncompressed and compressed data. Generally, the planned works provide the
compression ratio near layout 100% which is not available in other methods. This
method is well suitable for all type of images like satellite and medical images. At
the same time, the compression speed of ORT can be improved by parallelizing the
algorithm through two steps. The steps are data division and recursive runs pipeline
for the purpose of parallelization.

4.2. Physical- next generation secure computing

The hybrid MPI-open MP strategy is another name of PHY-NGSC.27 The main aim
of this method on ORT is the improvement of compression speed in the encoding
process because it is very important in case of multimedia applications. The vital
aspect of hybrid MPI-open MP strategy is its parallel programming on its layered
structure. The layers will be arranged in the order of group of pictures (GOP),

2050007-7
 P. Girishwaingankar & S. M. Joshi

picture, slice, macro block and blocks. Initially, collection of GOP forms a sequence
and each GOP contain a number of adjacent frames. Frames will undergo the
division of slices which is the combination of 16
16 pixel groups (macro block)
de¯ned as a fragment of rows in an image. Then blocks (8
8 pixels) are the sub
division of macro block which describes the luminance and chrominance of the pixel.
by UPPSALA UNIVERSITY on 05/18/20. Re-use and distribution is strictly not permitted, except for Open Access articles.

In this block, the decoder processes the encoded video frame.

The result of the ORT-based RLE is in the form of a coded bit stream which is the
combination of motion vector data and other information. At the initial level, par-
allelization across the sequence leads to load imbalance as a complex task. There is no
inherent communication among the independent GOP's in the parallel algorithm. In
the second level, adjacent images are synchronized by assigning various tasks to each
Int. J. Image Grap. 2020.20. Downloaded from www.worldscientific.com

image. The macro block has many tiny tasks to make the parallelism as a most
important approach at the slice level. At the level of parallelism at GOP, incoming
bit streams are decomposed into GOP which are assigned to MPI process. The
dividing of GOP based on block and cyclic manner and the second one is similar to
the round robin fashion. The ¯rst and second method gets the ¯rst and secondary
GOP until no more GOP remains for the processing.
Multiple open MP threads are running inside the MPI process to process every
slice within an image in case of parallelism at slice level. The pictures may be loaded
into the shared memory after the process of synchronization at the image ends.
Initially, the GOP and MPI processes are initialized in the proposed compression
speed enhancing methods. The compression is performed by reducing the average
number of bits at each frame of GOP. Similarly, slices are also processed from the
¯rst one to the ¯nal piece. Actually, the simultaneous functioning of block and cyclic
GOP divisions improves the compression speed.
Start
Initialize MPI process, Frame
for each row in a frame
f
Perform ORT based compression in a parallel manner
g
for row 0 to row n
f
Process every rows
g
Obtain the compressed results
end

This is the pseudo code for the proposed compression speed enhancement method.
First, initialize the MPI process and frames. Because the video is divided into several
frames in the ¯rst step. That is why the compression process is performed for every
frame in a row by row manner. In this hybrid MPI-open MP method, all rows are

2050007-8
 The PHY-NGSC-Based ORT Run Length Encoding Scheme for Video Compression

encoded simultaneously instead of one after another. It means, the time duration
taken for the compression process automatically reduces. Hence, the compression
speed is much better in the planned work through the usage of hybrid MPI-open MP
strategy.
by UPPSALA UNIVERSITY on 05/18/20. Re-use and distribution is strictly not permitted, except for Open Access articles.

5. Result and Discussion

The proposed method is implemented in the working platform of Matlab R2016a
with win-64 bit. In our system, three videos are taken and each one of the video's size
is 256*256*3 for the implementation purpose and the corresponding results for before
and after compression is shown in the following diagrams. Table 1 shows the
Int. J. Image Grap. 2020.20. Downloaded from www.worldscientific.com

Table 1. Experimental parameters.

Number of videos 3
Frame size of each video 256*256*3
RAM 4 Gb
Frames for each video 92

Before Compression After Compression

Fig. 4. Frames of test video 1 before and after compression.

2050007-9
 P. Girishwaingankar & S. M. Joshi

Before Compression After Compression

by UPPSALA UNIVERSITY on 05/18/20. Re-use and distribution is strictly not permitted, except for Open Access articles.
Int. J. Image Grap. 2020.20. Downloaded from www.worldscientific.com

Fig. 5. Frames of test video 2 before and after compression.

parameters taken to perform the experimental analysis. Initially, the video clip is
divided into several frames for the sake of performing the compression process. Then
every frame undergoes compression process individually. After the completion of the
compression procedure, the resultant frames are ¯nally converted into a video.
Figures 4–6 represent with and without compression results for three videos. The
resultant performances of these test videos are shown in Figs. 7–11.
PSNR, compression ratio, compression time, dictionary size and bit wise com-
pressed video size are given in the following results. The linear methods like
CAVLC,21 CBAC22 and accordion-based modi¯ed RLE24 are taken for the com-
parison purpose.
Figure 7 represents the comparison result of peak signal to noise ratio with other
encoding techniques. Actually, the proposed method has high value for PSNR
whereas the hybrid method gives the second maximum value. At the same time, the
PSNR value of accordion is more or less same that of the CAVLC. CBAC shows
lower performance than other methods by means of PSNR.

2050007-10
 The PHY-NGSC-Based ORT Run Length Encoding Scheme for Video Compression

Before Compression After Compression

by UPPSALA UNIVERSITY on 05/18/20. Re-use and distribution is strictly not permitted, except for Open Access articles.
Int. J. Image Grap. 2020.20. Downloaded from www.worldscientific.com

Fig. 6. Frames of test video 3 before and after compression.

The dictionary size of bits for various techniques along with the proposed
method is displayed in Fig. 8. The dictionary-based compression method denotes
the common sequence of pixels to capture the symbol correlation. The lower bit size
of the dictionary is given by the proposed method than others taken for the
comparison.
The ratio between the uncompressed video size and the compressed one is termed
as compression ratio. The compression ratio in percentage is given in Fig. 9. The
proposed methods perform better by means of compression ratio. That is, proposed
technique provides 99.75% of compression ratio which is better than accordion-based
modi¯ed RLE (86.6%). But the performance of other encoding techniques is poor
when compared with accordion and ORT-based RLE method.
The video size after the compression process is given in Fig. 10. The proposed
method gives a better-compressed size of nearly 0:5
10 5 bits than others. Whereas,
the accordion provides the compression size of 0:8
10 5 . All the other methods
contain the compressed size values greater than 3:2
10 5 . Generally, the compressed

2050007-11
 P. Girishwaingankar & S. M. Joshi
by UPPSALA UNIVERSITY on 05/18/20. Re-use and distribution is strictly not permitted, except for Open Access articles.
Int. J. Image Grap. 2020.20. Downloaded from www.worldscientific.com

Fig. 7. Comparison of peak signal to noise ratio (PSNR).

video size should be low for better performance. Hence, the planned work is good by
means of compression size in bits.
The time taken for the completion of the compression process is compared in
Fig. 11. The compression time is very high in CAVLC, CBAC and accordion
methods when compared with the proposed work. Normally, ORT method lacks by
means of compression speed. That is why hybrid MP-open MPI technique is

Fig. 8. Comparison of dictionary size in bits.

2050007-12
 The PHY-NGSC-Based ORT Run Length Encoding Scheme for Video Compression
by UPPSALA UNIVERSITY on 05/18/20. Re-use and distribution is strictly not permitted, except for Open Access articles.
Int. J. Image Grap. 2020.20. Downloaded from www.worldscientific.com

Fig. 9. Comparison of compression ratio.

introduced along with ORT to speed up compression. Therefore, the compression

duration is very low in the proposed methodology than other encoding approaches.
Table 2 expresses the comparison of video information among accordion and
proposed method. Actually, video clip contains 92 frames for both accordion and
proposed method respectively. The related parameters such as original video size,
compressed size and resized value are compared in the above table. Hence, the
analysis of all comparison parameters conformed that the proposed methodology is
better than other run-length encoding techniques.

Fig. 10. Comparison of compressed video size.

2050007-13
 P. Girishwaingankar & S. M. Joshi
by UPPSALA UNIVERSITY on 05/18/20. Re-use and distribution is strictly not permitted, except for Open Access articles.
Int. J. Image Grap. 2020.20. Downloaded from www.worldscientific.com

Fig. 11. Comparison of compression time.

Table 2. Comparison of video detail.

Parameters Accordion based modi¯ed RLE ORTþ PHY-NGSC-based RLE

Video Clip 92 Frames 92 Frames
Original Size 5.29 Mb 5.29 Mb
Size After Resizing 1.32 Mb 1.30 Mb
Size After Compression 0.662 Mb 0.049 Mb
Total Compression (%) 86.6% 99.75%

6. Conclusion
In this paper, octonary repetition tree (ORT) based RLE with PHY-NGSC has
been proposed. In proposed work, tree structure used to represent duplicate bytes
instead of °ag or code word. Better compression ratio of about 99.75% has been
achieved in proposed method which is greater than other exiting methods. The
created tree is pruned for reducing the number of bytes. But, the performance of the
ORT method was not good by means of compression speed. With the help of hybrid
MPI-open MP strategy, compression speed improvement have been achieved
through parallelizing the algorithm by dividing the data and pipelining the recur-
sive runs. Finally, the proposed method was compared with other compression
methods to show the better performance of PHY-NGSC-based ORT run length
encoding. In future, we will further construct new features to enhance the calcu-
lation of video segmentation and classi¯cation of the domain block and we will
strive for proposing a more e®ective parallel fractal compression method for real-
time big video data processing.

2050007-14
 The PHY-NGSC-Based ORT Run Length Encoding Scheme for Video Compression

References
1. C. M. Agulhari, I. S. Bonatti and P. L. Peres, An Adaptive Run Length Encoding Method
for the Compression of Electrocardiograms, Vol. 35 (Medical Engineering and Physics,
2013), pp. 145–153.
2. A. H. Robinson and C. Cherry, \Results of a prototype television bandwidth compression
by UPPSALA UNIVERSITY on 05/18/20. Re-use and distribution is strictly not permitted, except for Open Access articles.

scheme," Proc. IEEE, Vol. 55, 1967, pp. 356–364.

3. E. H. Yang and L. Wang, \Joint optimization of run-length coding, Hu®man coding, and
quantization table with complete baseline JPEG decoder compatibility," IEEE Trans.
Image Process. 18(1), 63–74 (2009).
4. G. J. Sullivan and T. Wiegand, \Video compression-from concepts to the H. 264/AVC
standard," Proc. IEEE, Vol. 93, 2005, pp. 18–31.
5. F. Hartung and B. Girod, \Watermarking of uncompressed and compressed video,"
Signal Process. 66(3), 283–301 (1998).
Int. J. Image Grap. 2020.20. Downloaded from www.worldscientific.com

6. A. D. Gupte, B. Amrutur, M. M. Mehendale, A. V. Rao and M. Budagavi, \Memory

bandwidth and power reduction using lossy reference frame compression in video
encoding," IEEE Trans. Circuits Syst. Video Technol. 21(2), 225–230 (2011).
7. H. Cheng and X. Li, \Partial encryption of compressed images and videos," IEEE Trans.
Signal Process. 48(8), 2439–2451 (2000).
8. X. Zhang, \Lossy compression and iterative reconstruction for encrypted image," IEEE
Trans. Inform. Forensics Security 6(1), 53–58 (2011).
9. T. Wieg and G. J. Sullivan, G. Bjontegaard and A. Luthra, \Overview of the H. 264/AVC
video coding standard," IEEE Trans. Circuits Syst. Video Technol. 13(7), 560–576
(2003).
10. S. Khan and M. Shah, \Object based segmentation of video using color, motion and
spatial information," Proc. Computer Society, Vol. 2, II–II (2001).
11. M. H. Pinson and S. Wolf, \A new standardized method for objectively measuring video
quality," IEEE Trans. Broadcasting 50(3), 312–322 (2004).
12. U. Gargi, R. Kasturi and S. H. Strayer, \Performance characterization of video-shot-
change detection methods," IEEE Trans. Circuits Syst. Video Technol. 10(1), 1–13
(2000).
13. C. Y. Lin and S. F. Chang, \A robust image authentication method distinguishing JPEG
compression from malicious manipulation," IEEE Trans. Circuits Syst. Video Technol.
11(2), 153–168 (2001).
14. S. G. Miaou, F. S. Ke and S. C. Chen, \A lossless compression method for medical image
sequences using JPEG-LS and inter frame coding," IEEE Trans. Inform. Technol.
Biomed. 13(5), 818–821 (2009).
15. A. Skodras, C. Christopoulos and T. Ebrahimi, \The JPEG 2000 still image compression
standard," IEEE Signal Process. Mag. 18(5), 36–58 (2001).
16. W. Lin and L. Dong, \Adaptive down sampling to improve image compression at low bit
rates," IEEE Trans. Image Process. 15(9), 2513–2521 (2006).
17. E. Elharar, A. Stern, O. Hadar and B. Javidi, \A hybrid compression method for integral
images using discrete wavelet transform and discrete cosine transform," J. Display
Technol. 3(3), 321–325 (2007).
18. J. Tang, J. Kim and E. Peli, \Image enhancement in the JPEG domain for people with
vision impairment," IEEE Trans. Biomed. Eng. 51(11), 2013–2023 (2004).
19. D. T. Vo, T. Q. Nguyen, S. Yea and A. Vetro, \Adaptive fuzzy ¯ltering for
artifact reduction in compressed images and videos," IEEE Trans. Image Process. 18(6),
1166–1178 (2009).

2050007-15
 P. Girishwaingankar & S. M. Joshi

20. T. Komatsu, Y. Ueda and T. Saito, \Super-resolution decoding of JPEG-compressed

image data with the shrinkage in the redundant DCT domain," Picture Coding Symp.
(PCS), IEEE, 2010, pp. 114–117.
21. G. C. Harisha, N. Murugendrappa, B. S. Manjushree, C. M. Kunjana, C. L. Sai and B. L.
Mahima, \Implementation of context adaptive variable length coding and decoding
algorithm for h. 264 video codec using Matlab," Int. J. Adv. Eng. Technol. 11(3), 113–121
by UPPSALA UNIVERSITY on 05/18/20. Re-use and distribution is strictly not permitted, except for Open Access articles.

(2018).
22. W. Yang, Y. Lin, S. Wu and R. Yu, \Improving coding e±ciency of MPEG-G standard
using context-based arithmetic coding," 2018 IEEE Int. Conf. Bioinformatics and Bio-
medicine (BIBM), December 2018, pp. 1177–1183.
23. L. Kong and R. Dai, \Object-detection-based video compression for wireless surveillance
systems," IEEE Multi Media 24(2), 76–85 (2017).
24. S. Ahmadand, M. M. Siddiqui, Video Compression Using Accordian Discrete Cosine
Int. J. Image Grap. 2020.20. Downloaded from www.worldscientific.com

Transform Method.
25. X. Zhang, R. Xiong, W. Lin, J. Zhang, S. Wang, S. Ma and W. Gao, \Low-rank-based
nonlocal adaptive loop ¯lter for high-e±ciency video compression," IEEE Trans. Syst.
Video Technol. 27(10), 2177–2188 (2017).
26. K. G. Haghighi, M. Mirniaand and A. H. Navin, Optimizing run-length algorithm using
octonary repetition tree, arXiv: 1611.09664.
27. S. Vargas-Perez and F. Saeed, \A hybrid MPI-Open MP strategy to speed up the com-
pression of big next-generation sequencing datasets," IEEE Trans. Parallel Distribute.
Syst. 28(10), 2760–2769 (2017).

Poorva Girishwaingankar is currently working as Associate

Professor in Electronics Engineering Department at Thakur Col-
lege of Engg. and Technology, Mumbai, India and pursuing Ph.D.
in Electronics Faculty at Swami Ramanand Tirth Marathwada
University, Nanded, India.

Sangeeta Milind Joshi is currently Professor and Technical

Advisor at Vidyalankar Institute of Technology, Mumbai, India.
She has published number of research papers in reputed journals
and received many awards. Her major ¯elds of interest are, Em-
bedded Systems and PC Based Instrumentation, VLSI Design,
Recon¯gurable Computing, Sensor Network, Simulation and
Modeling of Nanodevices.

2050007-16