AVC based Compression of Compound Images Using Block Classification Scheme

IJIS Print ISSN: 1911 8408
P.S.Jagadeesh Kumar et al, International Journal of Information Studies, Vol. 6 (4), 2015, 3479-3488
AVC based Compression of Compound Images
Using Block Classification Scheme
Dr.P.S Jagadeesh Kumar1
Professor
School of Computer Science and Engineering
Nanyang Technological University, Singapore
Dr.Krishna Moorthy2
Professor
Department of Computer Science and Engineering
Anna University, Chennai
Abstract-Most of the internet sites involve buffering of
compound images and people expect them to be of enhanced
quality and stature. In this paper, H.264/MPEG-4 AVC for
compound image compression using histogram based block
classification scheme is implemented to evaluate the
performance of compound images to better know the
importance of this scheme. In this method, conventional
histogram is adopted and improves the block classification
between background, text, hybrid and picture image areas.
Different compression scheme is implemented for different
types of classified blocks to improve the compression ratio to
an optimal value. H.264 AVC using CABAC entropy coding is
used to compress the hybrid blocks, on the other hand
standard conventional algorithm such as run-length, wavelet
coding and JPEG coding for background, text and picture
blocks respectively. Experimental results show that the
proposed scheme improves the compression ratio of
compound image to larger extent but increases the coding
complexity.
General Terms-Digital Image Compression, H.264 Advanced
Video Coding, CABAC Entropy Coding.
Keywords-Compound Image, Discrete Wavelet Transform,
Run-length coding, JPEG Coding, Histogram.
1. INTRODUCTION
With the widespread of digital devices such as digital
cameras, personal computers, more and more compound
images, containing text, graphics and natural images, are
available in digital forms such as screen images, web pages.
A compound image is an image that contains data of
various types such as photographic images, text and
graphics. Each of these data types has different statistical
properties, and is characterized by different level of
distortion that a human observer can notice. The sensitivity
of human eyes for natural image and text is different. The
quality requirement of compound image coding is different
from general image coding because users cannot accept the
quality if text is not clear enough to recognize. Most of the
efforts in image compression until now have been in
developing new algorithms that achieve better compression
at the cost of considerable increase in complexity. The
block classification algorithm and compression algorithm
have low calculation complexity, which makes them very
suitable for real time application. With the fast
development of internet and widespread rich media
applications, compound images such as web pages, slides
and posters are most commonly used. For natural images,
the existing image and video coding standards (e.g.,
JPEG2000 and H.264/AVC) have shown good coding
performance. However, they are not good at compressing
the compound images [1]. Thus, it is necessary to have
more research work on compound image compression.
2. COMPOUND IMAGE
Electronic document images often contain mixed content
types like, text, background, graphics in both gray scale and
in color form because of this mixed content, and they are
termed as “Compound Images”. It contains graphics, text
and natural images. A compound image occurs in many
important applications like document imaging and printing.
Sample compound image is shown in Fig 1. Segmentation
subdivides an image into its constituent regions or objects.
Image segmentation is typically used to locate objects and
boundaries (lines, curves, etc.) in images [1]. A general
approach to compress the compound image includes the
image segmentation into the regions of similar data types.
Bandwidth is a very important limiting factor in application
of image segmentation. Several segmentation schemes
require morphological analysis of the different regions, and
multiple passes over the image being segmented. However,
each pass normally requires loading memory data from
slow to fast memory (L1 cache, etc.), which is a slow
process.
Segmentation solutions based on multiple passes are much
slower (or costly) than what can be expected by, for
instance, counting the number of operations. Thus, an ideal
solution would use a single pass to decide on the type of
image region. Such solution would be very difficult with

arbitrary shapes of segmentation regions, but it are feasible
if considered only a predefined shape.
Fig 1: Sample Compound Image
3. BLOCK BASEDCOMPRESSION
The framework of the block-based compression
scheme is shown in Fig.2. The compound image is first
divided into 8x8 blocks. Then blocks are classified into
four types: background, text, mixed and picture according
to their different statistical characteristics. Blocks of
different type will be compressed with different algorithms.
The proposed scheme can effectively compress the mixed
blocks, which are not well handled by some block-based
algorithms. The proposed scheme achieves good coding
performance on text images, picture images and compound
images. It also outperforms DjVu on compound images at
high bit rate. The block type map is compressed using an
arithmetic coder [2].
Fig 2: Block Based Compression Scheme
4. BLOCK CLASSIFICATIONSCHEME
Block classification is defined as to classify the blocks into
individual blocks. Classification is performed by using the
histogram values. A histogram is a graphical representation,
showing a visual impression of distribution of experimental
data. Based upon the histogram, we have to set the
threshold values. The threshold values classify the blocks
as text, picture, background and mixed blocks. A fast and
effective classification algorithm based on three features:
histogram, gradient of the block and the
anisotropic values. The entire block classification flow is
shown in Fig.3. Blocks are classified into four types:
background, text, mixed and picture. Blocks of different
type are distinct in nature and have different statistics
properties.
The background blocks contain only the low histogram
pixels and show one peak at the low histogram pixels. The
text blocks always shows several peaks in low histogram
value (LHV). Only a few mid-histogram values (MHV) are
observed in text blocks. If the block contains large numbers
of high histogram and mid histogram values, it will be
identified as mixed blocks [2]. The block mainly consisting
of mid histogram values are declared as picture blocks.
Here thresholds T1-T4 is chosen to determine the block
type.
Fig 3: Block Classification Scheme
5. RUN LENGTHENCODING
Run-length encoding (RLE) is a very simple form of data
compression in which runs of data (that is, sequences in
which the same data value occurs in many consecutive data
elements) are stored as a single data value and count, rather
than as the original run. This is most useful on data that
contains many such runs. Consider, for example, simple
graphic images such as icons, line drawings, and
animations. It is not useful with files that don't have many
runs as it could greatly increase the file size.
RLE may also be used to refer to an early graphics file
format supported by CompuServe for compressing black
and white images, but was widely supplanted by their
later Graphics Interchange Format. RLE also refers to a
little-used image format in Windows 3.x, with the extension
rle, which is a Run Length Encoded Bitmap, used to
compress the Windows 3.x startup screen. Typical
applications of this encoding are when the source
information comprises long substrings of the same
character/binary digit. Run-length encoding performs
lossless data compression and well suited

to palette-based bitmapped images such as computer. It
does not work well at all on continuous-tone images such
as photographs; JPEG uses it quite effectively on the
coefficients that remain after transforming
and quantizing image blocks. Common formats for run-
length encoded data include Truevision TGA, Pack Bits,
PCX and ILBM. ITU also describes a standard to encode
run-length-color for fax machines, known as T.45. Run-
length encoding is used in fax machines (combined with
other techniques into Modified Huffman coding). It is
relatively efficient because most faxed documents are
generally white space, with occasional interruptions of
black.
For example, consider a screen containing plain black text
on a solid white background. There will be many long runs
of white pixels in the blank space, and many short runs of
black pixels within the text. A hypothetical scan line, with
B representing a black pixel and W representing white,
might read as follows:
WWWWWWWWWWWWBWWWWWWWWWW
WWBBBWWWWWWWWWWWWWWWWWWW
WWWWWBWWWWWWWWWWWWWW
With a run-length encoding (RLE) data compression
algorithm applied to the above hypothetical scan line; it can
be rendered as follows:
This can be interpreted as a sequence of twelve Ws, one B,
twelve Ws, three Bs, etc. The run-length code represents
the original 67 characters in only 18. While the actual
format used for the storage of images is generally binary
rather than ASCII characters like this, the principle remains
the same. Even binary data files can be compressed with
this method; file format specifications often dictate
repeated bytes in files as padding space. However, newer
compression methods such as DEFLATE often use LZ77-
based algorithms, a generalization of run-length encoding
that can take advantage of runs of strings of characters
(such as BWWBWWBWWBWW ).
Run-length encoding can be expressed in multiple ways to
accommodate data properties as well as additional
compression algorithms. For instance, one popular method
encodes run lengths for runs of two or more characters
only, using an "escape" symbol to identify runs, or using
the character itself as the escape, so that any time a
character appears twice it denotes a run [2]. On the
previous example, this would give the following:
This would be interpreted as a run of twelve Ws, a B, a run
of twelve Ws, a run of three Bs, etc. In data where runs are
less frequent, this can significantly improve the
compression rate.
One other matter is the application of additional
compression algorithms. Even with the runs extracted, the
frequencies of different characters may be large, allowing
for further compression; however, if the run lengths are
written in the file in the locations where the runs occurred,
the presence of these numbers interrupts the normal flow
and makes it harder to compress. To overcome this, some
run-length encoders separate the data and escape symbols
from the run lengths, so that the two can be handled
independently. For the example data, this would result in
two outputs, the string " WWBWWBBWWBWW " and
thus 12, 12, 3, 24, 14 .
6. WAVELET CODING
In wavelet image compression, after the quantization step
entropy encoding, which is a lossless form of compression
is performed on a particular image for more efficient
storage. Either 8 bits or 16 bits are required to store a pixel
on a digital image. Efficient entropy encoding, can use a
smaller number of bits to represent a pixel in an image; this
results in less memory usage to store or even transmit an
image. Karhunen-Lo`eve theorem enables us to pick the
best basis thus to minimize the entropy and error, to better
represent an image for optimal storage or transmission.
Also, Shannon-Fano entropy, Huffman coding,
Kolmogorov entropy and arithmetic coding are ones that
are used by engineers. Here, optimal means it uses least
memory space to represent the data. i.e., instead of using 16
bits, use 11 bits. Thus, the best basis found would make it
possible to represent the digital image with less storage
memory. In addition, the choices made for entropy
encoding varies; one might take into account the
effectiveness of the coding and the degree of difficulty of
implementation step into programming codes [3].
Fig 4: Wavelet Coding Scheme
7. JPEG CODING
JPEG Standard is the very well-known ISO/ITU-T standard
created in the late 1980s. JPEG standard is targeted for full-
color still frame applications. One of the most common
compression standards is the JPEG standard. Several modes
are defined for JPEG including baseline, lossless,
progressive and hierarchical. The most common mode uses
the discrete cosine transform is the JPEG baseline coding
system, also it is suitable for most compression
applications. Despite being developed for low
compressions JPEG it is very helpful for DCT quantization
and compression. JPEG compression reduces file size with
minimum image degradation by eliminating the least
important information. But it is considered a lossy image
compression technique because the final image and the
original image are not completely the same and in lossy
compression the information that may be lost and missed is
affordable. JPEG compression is performed in sequential
steps. JPEG compression is used in a number of image file
formats. JPEG/Exif is the most common image format used
by digital cameras and other photographic image capture
devices; along with JPEG/JFIF, it is the most common
format for storing and transmitting photographic images on
the World Wide Web. These format variations are often not
12W1B12W3B24W1B14W
WW12BWW12BB3WW24BWW14

distinguished, and are simply called JPEG. The term
"JPEG" is an acronym for the Joint Photographic Experts
Group, which created the standard. The MIME media type
for JPEG is image/jpeg (defined in RFC 1341), except in
older Internet Explorer versions, which provides a MIME
type of image/pjpeg when uploading JPEG images.
JPEG/JFIF supports a maximum image size of
65535×65535 pixels, hence up to 4 gigapixels (for an
aspect ratio of 1:1). JPEG uses a lossy form of compression
based on the discrete cosine transform (DCT). This
mathematical operation converts each frame/field of the
video source from the spatial (2D) domain into the
frequency domain (a.k.a. transform domain.) A perceptual
model based loosely on the human psychovisual system
discards high-frequency information, i.e. sharp transitions
in intensity, and color hue. In the transform domain, the
process of reducing information is called quantization. In
simpler terms, quantization is a method for optimally
reducing a large number scale (with different occurrences
of each number) into a smaller one, and the transform-
domain is a convenient representation of the image because
the high-frequency coefficients, which contribute less to the
overall picture than other coefficients, are characteristically
small-values with high compressibility. The quantized
coefficients are then sequenced and losslessly packed into
the output bitstream. Nearly all software implementations
of JPEG permit user control over the compression-ratio (as
well as other optional parameters), allowing the user to
trade off picture-quality for smaller file size. In embedded
applications (such as miniDV, which uses a similar DCT-
compression scheme), the parameters are pre-selected and
fixed for the application.
The compression method is usually lossy, meaning that
some original image information is lost and cannot be
restored, possibly affecting image quality. There is an
optional lossless mode defined in the JPEG standard.
However, this mode is not widely supported in products.
Widespread use of the format has stimulated the adoption
of simulated high-dynamic-range imaging (HDR) modes in
inexpensive cameras and smartphones, to correct the loss of
shadow and highlight detail [4,8].
There is also an interlaced "Progressive JPEG" format, in
which data is compressed in multiple passes of
progressively higher detail [3]. This is ideal for large
images that will be displayed while downloading over a
slow connection, allowing a reasonable preview after
receiving only a portion of the data. However, support for
progressive JPEGs is not universal. When progressive
JPEGs are received by programs that do not support them
(such as versions of Internet Explorer before Windows7)
the software displays the image only after it has been
completely downloaded. There are also many medical
imaging and traffic systems that create and process 12-bit
JPEG images, normally grayscale images.
The 12-bit JPEG format has been part of the JPEG
specification for some time, but this format is not as widely
supported. JPEG compression artifacts blend well into
photographs with detailed non-uniform textures, allowing
higher compression ratios. Notice how a higher
compression ratio first affects the high-frequency textures
in the upper-left corner of the image and how the
contrasting lines become more fuzzy. The very high
compression ratio severely affects the quality of the image,
although the overall colors and image form are still
recognizable. However, the precision of colors suffer less
(for a human eye) than the precision of contours (based on
luminance). This justifies the fact that images should be
first transformed in a color model separating the luminance
from the chromatic information, before sub sampling the
chromatic planes (which may also use lower quality
quantization) in order to preserve the precision of the
luminance plane with more information bits.
8. H.264 ADVANCED VIDEOCODING
In 2001, with the aim of developing a more efficient
compression system, the standardization bodies ISO/IEC
(MPEG) and the ITU brought their efforts together in the
Joint Video Team (JVT), a working group charged with
developing a coding system called Advanced Video Coding
(AVC). In 2003, the AVC system was integrated as part 10
in the MPEG-4 ISO/IEC 14496-10 standard and assumed
the name H.264 in the ITU. In September 2004, the DVB
Consortium modified ETSI standard TS 101 154 2
(Implementation guidelines for the use of Video and Audio
Coding in Broadcasting Applications based on the MPEG-2
Transport Stream) to also include AVC/H.264. The goal
behind the H.264 standard was to provide high quality
video at considerably lower bit rates than previous
standards. At the same time, the design needed to be not too
complex or expensive to implement. A secondary goal was
to make the standard flexible enough to be implemented
across a variety of applications, networks and systems.
Video standards have hit a sweet-spot based on the
compression and decompression technologies available
currently [5]. Because technology is roughly doubling in
performance every 18 months, every few years the industry
is able to make a quantum leap in performance. H.264
vastly improves compression performance over standards
such as MPEG-2 and MPEG-4 Visual. In the last few years,
many promising proprietary technologies (Windows Media,
Real Video, On2, Sorenson etc.) have been developed to
distribute multimedia content over narrow-bandwidth
transmission channels. These compression systems have
been used primarily for low-bitrate video at smaller picture
sizes (e.g. SQCIF, QCIF and CIF). However, their scope
has more recently been expanded to provide coding
efficiency across a wide range of bitrates – from a few
kbit/s up to tens of Mbit/s that can cope with SDTV and
HDTV.
9. CABAC ENTROPYCODING
Although Q-coder, MQ coder, and QM coder are also
binary arithmetic coders with statistics adaptivity features,
the Context-based Adaptive Binary Arithmetic Coder
(CABAC) – proposed by Marpe et al. in 2001 – is adopted
as the entropy coding tool for the Main profile and higher
profiles of the H.264/AVC standard. Before CABAC, LUT-
based VLC are generally utilized for the entropy coding
stage in the hybrid block-based video coding

standards including H.263, MPEG-2, MPEG-4 Part 2. The
weakness of VLCs is that coding event with probability
larger than 0.5 cannot be efficiently represented, and the
coding procedure is not adaptive to the actual symbol
statistics as the values of LUTs are fixed. The only
arithmetic coder adopted in video standard is of Annex E of
H.263 [ITU-T Rec. H.263], in which coding efficiency of
entropy coding is not significantly improved, because it
directly uses SEs of VLC for arithmetic coding without
redefinition. Before CABAC proposals by H.264/AVC,
similar arithmetic-coder-based implementations were first
investigated and applied to the non-block-based video
coding, such as DWT. CABAC is the first successful
arithmetic coding scheme deployed in the video coding
standard, with significant compression improvement
compared to the previous entropy coding tools. CABAC
encoding process consists of three elementary steps:
binarization, context modeling, and binary arithmetic
coding (BAC). The input SEs are binarized into bin strings,
in which regular bins and bypass bins are encoded
separately by the encoding engines of BAC. For regular bin
coding, the context model (probability model) of the bin is
prepared by the step of context modeling. Details of the
three steps are discussed in the following sections [5, 6].
9.1 Binarization
Binarization maps the non-binary valued SE into bin string,
which is a sequence of binary decision (bin). Three types of
bins are generated in the binarization step: regular bin,
bypass bin, and terminate bin for the bins with unequal
(variable), equal, or dominant probabilities of value 1 and
0, respectively. Advantages of binarization include:
(a) The probability of non-binary SE can be represented by
the probabilities of individual coding bins, while
compression efficiency is not influenced;
(b) Low-complexity BAC can be utilized;
(c) Context modeling at sub-symbol (sub-SE) level
provides more accurate probability estimation than context
modeling at the symbol level, and the alphabet of the
encoder is reduced. Five binarization schemes are used in
CABAC: Unary (U), Truncated Unary (TU), kth order Exp-
Golomb (EGk), concatenation of the first and third scheme
(UEGk), and fixed length binarization (FL). Kth ordered
Exp-Golomb binarization (EGk) – a derivative of Golomb
coding is proved to be an optimal prefix-free coding for
geometrically distributed sources. EGk codeword consists
of prefix and suffix bin strings, with total length of 2l+k+1
bits. EGk prefix is a Unary codeword, with l bits of 1 and
one terminating bit 0. The length l of string of bit 1 is
represented as:
The length of suffix binary string is equal to l + k, and the
value of the suffix string is:
UEGk is combined binarization scheme of TU and EGk. It
is utilized for binarization of SEs of absolute value of
residual coefficient level and MVD. TU generates prefix of
the bin string, and EGk is adopted to generate the suffix
with k set to 0 and 3 for coefficient level and MVD,
respectively. TU is simple and it permits fast adaptation of
probability of coding symbol. However, it is only beneficial
for small SE values. For large SE values, suffix bin string
generated by EGk provides a good fit to the probability
distribution, and bypass bin coding is also utilized to reduce
computation complexity.
9.2 Context modeling
The context modeling implements two sub-functions:
context model selection and context model access. The
statistics of the coded SEs are utilized to update the
probability models (context model) of regular bins. For
regular bin coding, one context model is chosen, and
fetched from a pre-defined set of context models to provide
the probability of regular bin to be MPS or LPS, and the
context model is updated after bin coding based on bin
value. The context index (CtxIdx) is calculated to select
context model, which is the sum of context offset
(CtxOffset) and context index increment (CtxIdxInc).
CtxOffset locates the context model set of processed SE,
while CtxIdxInc selects one context model from the set
based on the values of coded bins or coded SEs of
neighboring coded blocks. The idea of multiplication-free
arithmetic coding of H.264/AVC is based on the
assumption that estimated probability of each context
model can be represented by a sufficiently limited set of
representative values. In CABAC, the number of the
representative values is set to 64 to enable accurate
estimation, which is larger than the 30 of Q-coder. Each
context model contains a 1-bit tag of MPS value, and a 6-
bit probability state index (pStateIdx) that addresses one of
64 representative probability values of LPS from p0 to p63
in the range of [0.01875, 0.5]. The ratio of two neighboring
probability values is a constant value α, which is
approximately equal to 0.949.
By mapping the update probability value of LPS of (2.4) to
the closest value in the aforementioned set of representative
values, multiplication of probability estimation of CABAC
is replaced by simple table lookup for the pStateIdx of next
probability state according to pStateIdx of the current bin,
and based on whether it is MPS or LPS. This probability
value estimation of context state is actually the function
state transition of FSM with 64 predefined states. This type
of probability FSM is first utilized in Q-coder, and adopted
in QM coder and MQ coder. Compared to Q-coder, QM
coder, and MQ coder, the representative LPS probability
values need not to be stored in CABAC. Instead, the
approximation of the products of coding interval Range and
the LPS probability of (1.37) are stored. In order to be more
adaptive to the coding context, the values of MPS and LPS
can be exchanged when the probabilities of MPS and LPS
are equal and the coding bin is LPS.
For particular regular bins of CABAC, multiple context
models are allocated for single bin to more precisely
represent probabilities of bin in different coding contexts.
Four types of context model selection techniques are

supported in CABAC, based on (a) neighboring coded SE
values of the current SE, (b) values of prior coded bins of
SE bin string, (c) position of the to-be-encoded residual
coefficient in the scanning path of residual block
coefficients, and (d) level values of encoded coefficients of
residual block. If the coding bin is MPS, the probability of
LPS decreases by simply multiplying the ratio α, while for
the LPS bin, the update probability of MPS is calculated
first, and then the probability of LPS are obtained.
9.3 Binary Arithmetic Coding (BAC)
BAC performs arithmetic coding of each bin based on bin
value, type, and the corresponding context model of the
bin. BAC is a recursive procedure of coding interval
subdivision and selection. Coding interval subdivision
mechanism of CABAC is different from that of QM and
MQ coders. In QM and MQ coders, calculation of
RangeLPS is simplified by using the approximated value 1
for Range, and multiplication is eliminated. In comparison,
Range is also utilized for RangeLPS calculation of
CABAC. The reference interval subdivision and selection
of regular bin, in which the 2 higher-order bits of Range
(bit 7 and bit 6) and the index of probability state
(pStateIdx) of LPS are used to look up for the precalculated
product of Range and pLPS in a 2-D LUT. Although the
product of LUT is of limited precision, the precision of
RangeLPS calculation, interval subdivision is improved,
and computational complexity is reduced in CABAC,
compared to those of QM and MQ coders. Because Range
and Low of coding interval are represented by finite
number of bits (9 bits for Range and 10 bits for Low), it is
necessary to renormalize (scale up) the interval to prevent
precision degradation, and the upper bits of Low are output
as coded bits during renormalization.
Fig 5: CABAC Entropy Coding
Coding interval renormalization and bit output of CABAC
is based on Witten‟s algorithm. The coding interval of
(Low, Low + Range) is renormalized when Range is
smaller than the threshold value 256 (0x100), which is 1/4
of the maximum range of coding interval. Renormalization
of Range and Low is an iterative procedure, and the
maximum number of iterations is 6, as the smallest possible
value of Range is 6. For the processing of carry
propagation and output of coding bits, the coded bits of
CABAC are not output until it is confirmed that further
carry propagation will not influence bit values. Only when
interval length (Range) is smaller than the threshold 0x100,
one bit can be output if the interval is located within the top
half [0x200, 0x400) or bottom half [0, 0x200) of the
maximum coding range, or an outstanding (OS) bit is
accumulated when the interval is within [0x100, 0x300).
When a bit of value X is output in BAC, the accumulated
OS bits are output with the value 1–X. Compared to the bit
stuffing or byte stuffing schemes of Q-coder, QM coder,
and MQ coder, carry propagation is completely solved
during renormalization of BAC, and no additional
processing of bit stream is needed at CABAC decoder.
Moreover, as no bits or bytes are stuffed in the bit stream,
the compression efficiency of CABAC is further improved.
However, the renormalization illustrated in Fig. 5 is a
highly sequential operation, and as the number of iterations
is variable depending on the selected subinterval Range, it
is challenging for SW or HW acceleration of
renormalization and bit output of BAC. In some situations,
long delay may be experienced, when a large number of OS
bits are accumulated.
10. IMPLEMENTATION
In this compression scheme, the compound image is given
as input and then converted to grayscale. The grayscale
array image is further classified into background, text,
hybrid and natural image using histogram based block
classification method. The classified blocks are further
compressed using different compressed algorithm. H.264
AVC using CABAC entropy coding is used to compress
the hybrid blocks, on the other hand standard conventional
algorithm such as run-length, wavelet coding and JPEG
coding for background, text and picture blocks
respectively.
10.1 Getting the input compound image and
converting it to grayscale
The compound image is given as input. The input
compound image is converted to grayscale. Calculates the
size of the input compound image and returns the number
of rows and columns of the input compound image pixel
array.
10.2 Using Discrete Wavelet Transform to get the
transformed values
Daubechies 9/7 which is used here is usually used for
lossless compression. In general Daubechies wavelet has
extremal phase and highest number of vanishing moments
for defined support width. The wavelet is also easy to put
into practice with minimum-phase filters. This wavelet is
often also called the CDF 9/7 wavelet (where 9 and 7
denote the number of filter taps). There are several ways

wavelet transforms can decompose a signal into various sub
bands. These include uniform decomposition, octave-band
decomposition, and adaptive/wavelet packet
decomposition. Out of these, octave-band decomposition is
the most widely used. The decomposition of the signal into
different frequency bands is simply obtained by successive
high pass and low pass filtering of the time domain signal.
This filter pair is called the analysis filter pair. First, the
low pass filter is applied for each row of data, thereby
getting the low frequency components of the row. But since
the low pass filter is a half band filter, the output data
contains frequencies only in the first half of the original
frequency range. They are down-sampled by two, so that
the output data contains only half the original number of
samples. Now, the high pass filter is applied for the same
row of data, and similarly the high pass components are
separated.
10.3 Encoding and Decoding Phase
The sample values are encoded and then decoded using
Context adaptive base arithmetic coding (CABAC). When
entropy coding mode is set to 1, an arithmetic coding
system is used to encode and decode H.264 syntax
elements. The arithmetic coding scheme selected for H.264,
Context-based Adaptive Binary Arithmetic Coding or
CABAC, achieves good compression performance through
(a) Selecting probability models for each syntax element
according to the element‟s context (b) Adapting probability
estimates based on local statistics and (c) using arithmetic
coding.
Coding a data symbol involves the following stages:
10.3.1. Binarization: CABAC uses Binary Arithmetic
Coding which means that only binary decisions (1 or 0) are
encoded. A non-binary-valued symbol (e.g. a transform
coefficient or motion vector) is “binarized” or converted
into a binary code prior to arithmetic coding. This process
is similar to the process of converting a data symbol into a
variable length code but the binary code is further encoded
(by the arithmetic coder) prior to transmission. Stages 2, 3
and 4 are repeated for each bit (or “bin”) of the binarized
symbol.
10.3.2. Context model selection: A “context model” is a
probability model for one or more bins of the binarized
symbol. This model may be chosen from a selection of
available models depending on the statistics of recently-
coded data symbols. The context model stores the
probability of each bin being “1” or “0”.
10.3.3. Arithmetic encoding: An arithmetic coder encodes
each bin according to the selected probability model. Note
that there are just two sub-ranges for each bin
(corresponding to “0” and “1”).
10.3.4. Probability update: The selected context model is
updated based on the actual coded value (e.g. if the bin
value was “1”, the frequency count of “1”s is increased).
10.3.5. Calculating compression ratios and evaluating
performance:
The compressed image pixel array is obtained from Inverse
DWT. Now from this signal to noise ratio is calculated
after calculating MSE. In order to evaluate the performance
of video compression coding, it is necessary to define a
measure to compare the original video and the video after
compressed. Most video compression systems are designed
to minimize the mean square error (MSE) between two
video sequences Ψ1 and Ψ2, which is defined as
11. COMPARISONS OF CABAC WITH OTHER ENTROPY
CODERS
The coding efficiency of CABAC is higher than those of
the Q-coder, QM coder, and MQ coder, because of (a) the
more precise multiplication of RangeLPS, (b) larger
number of probability states for each probability model,
and more precise probability estimation of coding bins; and
(c) more context models (probability models) deployed for
various coding contexts of different types of SEs. Because
of the high computational complexity of CABAC, another
entropy coding tool CAVLC is deployed in the Baseline
profile and extended profile of H.264/AVC targeting low
bit-rate real-time video coding. It offers compression-
complexity trade-off with lower complexity, and lower
coding efficiency, compared to CABAC. It is employed to
encode the quantized transform coefficients of 4×4 residual
blocks, while zero-order Exp-Golomb codes (EG0) are
used for all other types of non-residual SEs. Adaptivity is
introduced to CAVLC by allowing switching among
multiple VLC tables based on the already processed SEs,
and the coding efficiency of CAVLC is better than those of
the previous VLC coders which used single VLC table.
Instead of coding data pair of run-level as a single SE, run
and level of the residual block are encoded separately in
CAVLC, so that the inter-symbol redundancy can be more
efficiently exploited. However, compression efficiency of
CABAC is significantly higher, with typically bit rate
reduction of 9–14% in the video quality range of 30–38 dB,
compared to CAVLC and EG0. This is because (a) in
CABAC, encoding symbols can be more precisely
represented in non-integer number of bits, especially for the
symbol with probability higher than 0.5, and (b) CABAC
encoder is more adaptive to the non-stationary symbol
statistics with efficient context modelling (probability
estimation) for the coding bins of all types of SEs. Since
the adoption of CABAC entropy coding in H.264/AVC,
CABAC is also applied in many applications of image and
video processing including motion mode and residual data
of 3D dynamic mesh, prediction residual in lossless 4D
medical image compression, SEs of 8×8 transform
coefficients of AVS coding standard, motion vector coding
of scalable video coder, parameters of depth and correction
vectors in multi-view video coding. CABAC is also utilized
to encode affine motion vector, and MVD of 3-D DWT-
based subband video encoder [6, 7, 9].
12. EXPERIMENTAL RESULT
The famous toy store compound image as shown in Fig.6 is
taken as the input image to our proposed H.264/MPEG-4
AVC using histogram based block classification scheme.
The segmented image is shown in the Fig.7. The block
classification for text, hybrid, background and picture
blocks are shown in the Fig. 8, 9, 10, 11 respectively. The
proposed system has been simulated in MATLAB-
SIMULINK environment. H.264 AVC using CABAC
entropy coding is used to compress the hybrid blocks, on

the other hand standard conventional algorithm such as run-
length, wavelet coding and JPEG coding for background,
text and picture blocks respectively. The proposed scheme
is tested for different type of input compound images as
shown in Fig. 13, 15, 17, 19 and their corresponding
compressed images are shown in Fig. 14, 16, 18, 20.
Fig 6: Original input compound image
Fig 7: Converted grayscale image
Fig 8: Segmented image
Fig 9: Text classified blocks
Fig 10: Hybrid classified blocks
Fig 11: Background classified blocks
Fig 12: Picture classified blocks

Fig 13: Input compound image 1
Fig 14: Compressed compound image 1
Fig 18: Compressed compound image 3
Fig 16: Compressed compound image 2 Fig 20: Compressed compound image 4

CONCLUSION AND FUTURE ENHANCEMENT
The block classification scheme based on histogram
method is very simple and effective, reducing the
computational complexity. The histogram based block
classification scheme was 92% efficient for compound
images [1]. Considering the fact that sensitivity for human
eyes can mismatch 3% of block classification, though the
proposed scheme failed to make the same consistency for
other type of images as such can be negotiated. The
classified text blocks, background blocks, picture blocks
and hybrid blocks are compressed using wavelet coding,
run-length encoding, JPEG coding and H.264 /MPEG-4
AVC respectively. The PSNR and Compression ratio were
found to be comparatively better but the proposed scheme
is very heavy and complicated because of various
algorithms associated with it. In future, a single
compression scheme might reduce the coding complexity to
greater extent by implementing methods such as genetic
algorithm, neural network, sub-band coding.
REFERENCES
[1] P.S.Jagadeesh Kumar, “Pioneering VDT Image Compression using
Block Coding”, IEE Proceedings - Vision, Image and Signal
Processing, 152 (4), August 2005, pp. 513-518.
[2] Cuiling lan, Guangming Shi and Feng wu, “Compress Compound
Images in H.264/MPEG-4 AVC by exploiting Spatial Correlation”
IEEE Transactions on Image Processing, vol.19 no.4, pp. 946-957,
April 2010.
[3] P.S.Jagadeesh Kumar, Krishna Moorthy, “Intelligent Parallel
Processing and Compound Image Compression”, Advances in
Parallel Computing, New Frontiers in Computing and
Communications 2013, Vol. 38, Issue 1, January 2013, pp.196-205.
[4] D. Mukherjee, N. Memon, and A. Said, “JPEG-matched MRC
compression of compound documents,” Proc. IEEE Int. Conf. Image
Processing, vol. 3, pp. 434–437, Oct. 2001.
[5] P.S.Jagadeesh Kumar, J.Tisa, J.Lepika, J.Nedumaan, “Wavelet
Transform Based Reduction of Speckle in Ultrasound Images”,
International Journal of Imaging, 4 (3), June 2011, pp. 92-99.
[6] B.-f Wu, C.-C Chiu and Y. –L Chen, “Algorithms for compressing
compound document images with large text/background overlap”,
IEEE Proc.Vis. Image signal Process, Vol. 151 No. 6, pp.453-459
December 2008. Brown, L. D., Hua, H., and Gao, C. 2003. A widget
framework for augmented interaction in SCAPE.
[7] P.S.Jagadeesh Kumar, “Compression of Compound Images Using
Wavelet Transform”, Asian Journal of Computer Science and
Technology, 2 (2), 2013, pp. 263-268
[8] Wenpeng ding, Yan lu and Feng wu, “Enable Efficient Compound
Image Compression in H.264/AVC Intra Coding” IEEE Transactions
on Image Processing, Vol.10 no.3, pp. 337-340, Sep.2009.
[9] P.S.Jagadeesh Kumar, “„Efficient Block Classification of Computer
Screen Images for Desktop Sharing using Neural Network”, Journal
of Electrical and Computer Engineering, February 2013, Hindawi,
pp.703-711.
Corresponding Author
Dr.P.S.Jagadesh Kumar, Professor in the School of
Computer Science and Engg., Nanyang Technological
University, Singapore has sixteen years of teaching
experience, including 9 year of research experience in
the field of image compression. He received his B.E
degree from University of Madras in Electrical and
Electronics Engineering discipline in the year 1999. He
obtained his M.E degree in 2004 with specialization in
Computer Science and Engineering from Annamalai
University, Chidambaram and his doctorate from
University of Cambridge, United Kingdom.

AVC based Compression of Compound Images Using Block Classification Scheme

More Related Content

What's hot (20)

Similar to AVC based Compression of Compound Images Using Block Classification Scheme (20)

More from DR.P.S.JAGADEESH KUMAR (20)

Recently uploaded (20)

AVC based Compression of Compound Images Using Block Classification Scheme