Journal of Engineering Research and Studies E-ISSN0976-7916

Research Article
REAL TIME VIDEO COMPRESSION IMPLEMENTED USING
ADAPTIVE BLOCK TRANSFER / MOTION COMPENSATION
FOR LOWER BIT RATES
S.L. Lahudkar# and R.K. Prasad*
Address for Correspondence
# : Ph.D student , Bharati Vidyapeeth Deemed University, Pune
*: Guide; Head Dept. of Electronics & Telecommunication , AIT Pune
Affiliation: 1) Bharati Vidyapeeth Deemed University College of Engineering, Pune
2) JSPMs Imperial College of Engineering & Research, Pune
ABSTRACT
The aim of this paper is to achieve considerable amount of compression in real-time images by using adaptive block
transform with lower bit rates. The project aims to achieve the above mentioned objective by developing a code on a Turbo
C++ platform / MATLAB. The project discusses important features of adaptive block transform in compression of still
images, including the extent to which the quality of image is degraded by the process of discrete cosine transform
compression and decompression. Image quality is measured objectively, using peak signal-to-noise ratio or picture quality
scale, and subjectively, using perceived image quality. The effects of different functions, image contents and compression
ratios are assessed. A comparison with a discrete-cosine-transform-based compression system is given. Our results provide a
platform for reduced memory space and bandwidth requirements in on-line applications like video conferencing. Our project
also discusses the challenges faced while developing an efficient solution for bandwidth and memory problems on a Turbo
C++ platform. The still image formats used are essentially BMP (Bitmap), JPEG (Joint Photographic Experts Group)
and GIF (Graphical Inter-Face).On similar lines the real-time image formats we have worked with are MPEG(Motion
Picture Experts Group) and AVI(Audio-Visual Interface) .
KEY WORDS Compression, JPEG, Discrete Cosine Transform, Inverse Discrete Cosine Transform, Quantization, Run
Length Coding, Entropy, Adaptive Block Transfer.
1.0 INTRODUCTION: the sender's data more concisely, but nevertheless
1.1 Types of compression: There are two basic perfectly. Lossless compression is possible because
compression schemes: most real-world data has statistical redundancy. For
1.1.1: Lossless vs. lossy compression: The example, in English text, the letter 'e' is much more
advantage of lossy methods over lossless methods is common than the letter 'z', and the probability that the
that in some cases a lossy method can produce a letter 'q' will be followed by the letter 'z' is very
much smaller compressed file than any known small. Another kind of compression, called lossy data
lossless method, while still meeting the requirements compression, is possible if some loss of fidelity is
of the application. Lossy methods are most often used acceptable. For example, a person viewing a picture
for compressing sound, images or videos. The or television video scene might not notice if some of
compression ratio (that is, the size of the compressed its finest details are removed or not represented
file compared to that of the uncompressed file) of perfectly. Similarly, two clips of audio may be
lossy video codecs are nearly always far superior to perceived as the same to a listener even though one is
those of the audio and still-image equivalents. Audio missing details found in the other. [1] Lossy data
can be compressed at 10:1 with no noticeable loss of compression algorithms introduce relatively minor
quality, video can be compressed immensely with differences and represent the picture, video, or audio
little visible quality loss, eg 300:1. [5] Lossily using fewer bits. Lossless compression schemes are
compressed still images are often compressed to reversible so that the original data can be
1/10th their original size, as with audio, but the reconstructed, while lossy schemes accept some loss
quality loss is more noticeable, especially on closer of data in order to achieve higher compression.
inspection. When a user acquires a lossily- However, lossless data compression algorithms will
compressed file, (for example, to reduce download- always fail to compress some files; indeed, any
time) the retrieved file can be quite different from the compression algorithm will necessarily fail to
original at the bit level while being indistinguishable compress any data containing no discernible patterns.
to the human ear or eye for most practical purposes. Attempts to compress data that has been compressed
Many methods focus on the idiosyncrasies of the already will therefore usually result in an expansion,
human anatomy, taking into account, for example, as will attempts to compress encrypted data. In
that the human eye can see only certain frequencies practice, lossy data compression will also come to a
of light. The psycho-acoustic model describes how point where compressing again does not work,
sound can be highly compressed without degrading although an extremely lossy algorithm, which for
the perceived quality of the sound. Flaws caused by example always removes the last byte of a file, will
lossy compression that are noticeable to the human always compress a file up to the point where it is
eye or ear are known as compression artifacts.[5] empty.[2]
Lossless compression algorithms usually exploit 2.0 Theory:
statistical redundancy in such a way as to represent 2.1: What is JPEG?

JERS/Vol. II/ Issue III/July-September, 2011/147-153

 Journal of Engineering Research and Studies
"Joint Photographic Expert Group". Voted as
international standard in 1992. [4]
Works with color and grayscale images, e.g.,
satellite, medical, ...
2.2 Motivation
The compression ratio of lossless methods (e.g.,
Huffman, Arithmetic, LZW) is not high enough for
image and video compression, especially when the
distribution of pixel values is relatively flat. [4] JPEG
uses transform coding, it is largely based on the
following observations:
Observation 1: A large majority of useful image
contents change relatively slowly across images, i.e.,
it is unusual for intensity values to alter up and down
several times in a small area, for example, within an
8 x 8 image block. Translate this into the spatial
frequency domain, it says that, generally, lower
spatial frequency components contain more
information than the high frequency components
which often correspond to less useful details and
noises.[4]
2.3 JPEG overview
2.3.1 Encoding
Fig 1. JPEG Encoder
2.4 Major Steps for JPEG Encoding [4]
DCT (Discrete Cosine Transformation)
Quantization
Zigzag Scan
DPCM on DC component
RLE on AC Components
Entropy Coding
2.4.1 Discrete Cosine Transform (DCT) [1, 8]
Fig 2. Conversion from Spatial to Frequency
Domain
Discrete Cosine Transform (DCT):
Inverse Discrete Cosine Transform (IDCT):
JERS/Vol. II/ Issue III/July-September, 2011/147-153
E-ISSN0976-7916
Fig 3. Representation of Horizontal & Vertical
coefficients of DCT
Most software implementations use fixed point
arithmetic. Some fast implementations approximate
coefficients so all multiplies are shifts and adds.
2.4.2 Quantization [1,9]
F'[u, v] = round ( F[u, v] / q[u, v] ).
Quantization error is the main source of the
Lossy Compression.
Uniform Quantization
Each F[u,v] is divided by the same constant
N.
Non-uniform Quantization -- Quantization Tables
Eye is most sensitive to low frequencies
(upper left corner), less sensitive to high
frequencies (lower right corner)
Table 1.Luminance Quantization Table q(u, v)
Table 2.Chrominance Quantization Table q(u, v)
The numbers in the above quantization tables can be
scaled up (or down) to adjust the so called quality
factor.
Custom quantization tables can also be put in
image/scan header.
2.4.3 Zig-zag Scan [1,5]
Why? -- to group low frequency coefficients
in top of vector.
Maps 8 x 8 to a 1 x 64 vector
caled up (or down) to adjust the so called quality
factor.
 Journal of Engineering Research and Studies
Fig 4. Zig- Zag Scanning
2.4.4. Differential Pulse Code Modulation
(DPCM) on DC component
DC component is large and varied, but often
close to previous value.
Encode the difference from previous 8 x 8
blocks -- DPCM [5]
2.4.5. Run Length Encode (RLE) on AC
components
1 x 64 vector has lots of zeros in it
Keeps skip and value, where skip is the
number of zeros and value is the next non-
zero component.
Send (0,0) as end-of-block sentinel value.
[5]
2.4.6 Entropy Coding
Categorize DC values into SIZE (number of
bits needed to represent) and actual bits. [4]
Example: if DC value is 4, 3 bits are needed.
Send off SIZE as Huffman symbol, followed by
actual 3 bits.
For AC components two symbols are used:
Symbol_1: (skip, SIZE), Symbol_2: actual
bits. Symbol_1 (skip, SIZE) is encoded using
the Huffman coding, Symbol_2 is not
encoded.
Huffman Tables can be custom (sent in
header) or default.
Fig 5. JPEG Bit Stream
2.5 JPEG Modes
Lossless Mode
Sequential Mode
Progressive Mode
Hierarchical Mode
2.5.1 Lossless Mode
A special case of the JPEG where indeed there is no
loss. It does not use DCT-based method! Instead, it
uses a predictive(differential coding) method:
A predictor combines the values of up to three
neighboring pixels (not blocks as in the Sequential
mode) as the predicted value for the current pixel,
indicated by "X" in the figure below. The encoder
JERS/Vol. II/ Issue III/July-September, 2011/147-153
E-ISSN0976-7916
then compares this prediction with the actual pixel
value at the position "X", and encodes the difference
(prediction residual) losslessly.
It can use any one of the following seven
predictors.[2,4]
Table 3.Prediction table
Predictor Prediction
1 A
2 B
3 C
4 A+B-C
5 A + (B - C) / 2
6 B + (A - C) / 2
7 (A + B) / 2
Since it uses only previously encoded
neighbors, the very first pixel I(0, 0) will
have to use itself. Other pixels at the first
row always use P1, at the first column
always use P2.
Effect of Predictor (test result with 20
images):
Fig 6. Compression ratio at various Prediction
Levels
2.5.2 Sequential Mode
Each image component is encoded in a single left-to-
right, top-to-bottom scan.
Baseline Sequential Mode, the one that we described
above, is a simple case of the Sequential mode:
It supports only 8-bit images (not 12-bit images)
It uses only Huffman coding (not Arithmetic coding)
[2,4]
2.5.3. Progressive Mode
Goal: display low quality image and
successively improve.
Two ways to successively improve image:
1. Spectral selection: Send DC component
and first few AC coefficients first, then
gradually some more ACs. [3]
2. Successive approximation: send DCT
coefficients MSB (most significant bit)
to LSB (least significant bit).
Effectively, it is sending quantized DCT
coefficients frist, and then the
difference between the quantized and
the non-quantized coefficients with
finer quantization stepsize.[3]
2.5.4 Hierarchical Mode
(a) Down-sample by factors of 2 in each dimension,
e.g., reduce 640 x 480 to 320 x 240
(b) Code smaller image using another JPEG mode
(Progressive, Sequential, or Lossless).
 Journal of Engineering Research and Studies
Fig 7. Three level Hierarchical JPEG Encoder
(c) Decode and up-sample encoded image
(d) Encode difference between the up-sampled and
the original using Progressive, Sequential, or
Lossless.
Can be repeated multiple times.
Good for viewing high resolution image on
low resolution display.
3. 0 Methodology:
3.1 Transmitter
1. Implementation of RT Video
conferencing using DCT / ABT
(MATLAB)

Camera interfacing

Converting video data into frames
(JPEG)

JPEG encoding

Applying color space transformation

Down sampling

Block splitting

DCT using Adaptive Block Transfer

Quantization

Transmitting compressed data serially
through com port
3.2 Receiver

Receiving compressed data serially
through com port

De Quantization

Block splitting

IDCT

JPEG Decoder

Integrating frames to a video file using
3.3 Low Power Design of DCT / IDCT for Lower
Bit Rates:
Low bit rate wireless video systems have
applications in cellular videophones, wireless
surveillance systems.
Fig below shows the encoder and decoder
for lower bit rate systems.
Starting at the encoder the Motion Estimation
block(MEC) performs the temporal compression by
computing the difference between the current frame
and the previous frame . The temporal compressed
data is sent to DCT which is a key component in
spatial compression of data. The DCT block
transforms the data into spatial frequency
coefficients. The quantization block divides each
DCT coefficient by a quantization parameter setting
JERS/Vol. II/ Issue III/July-September, 2011/147-153
E-ISSN0976-7916
insignificant DCT coefficients to zero. The variable
length coder (VLC) performs run length coding on
the quantized coefficients compressing long runs of
DCT coefficients.
The coder sends the compressed data to the
decoder, which applies inverse process to reconstruct
the frame. The variable run length decoder(VLC-1)
reconstructs the quantized data and inverse
quantization block (Quant-1) block multiples by the
quantization parameter to recreate the DCT
coefficients. Next the IDCT block transforms the
DCT coefficients back to the spatial domain. Finally
the inverse estimation block. (MEC-1). [6]
Fig 8. Encoder / Decoder for low power & Low Bit
Rate
4.0 Innovations Implemented:
4.1 Adaptive Block Size Transform:
Traditionally, 8 8 transforms have been used for
image and video coding. However, there is evidence
to suggest that 4 4 transforms can reduce ringing
artifacts at edges and discontinuities. It is capable of
coding an 8 8 block using either an 8 8 transform,
two 8 4 transforms, two 4 8 transforms, or four 4
4 transforms. This feature enables coding that takes
advantage of the different transform sizes as needed
for optimal image quality.[7]
Fig 9. ABT applied to 8 x 8 Block with and
without Pre-filtering
4.2 Motion Compensation:
Motion compensation is the process of generating a
prediction of a video frame by displacing the
reference frame. Typically, the prediction is formed
for a block (an 8 8 pixel tile) or a macroblock (a 16
16 pixel tile) of data. The displacement of data due
to motion is defined by a motion vector, which
captures the shift along both the x- and y-axes.The
efficiency of the codec is affected by the size of the
predicted block, the granularity of sub-pixel data that
can be captured, and the type of filter used for
generating sub-pixel predictors. VC-1 uses 16 16
blocks for prediction, with the ability to generate
mixed frames of 16 16 and 8 8 blocks. Two sets
of filters are used by VC-1 for motion compensation.
The first is an approximate bicubic filter with four
taps. The second is a bilinear filter with two taps.
[7,10]
 Journal of Engineering Research and Studies E-ISSN0976-7916

5.0 Experimental Results:

Reference code
vid = videoinput('winvideo',1);
vidRes = get(vid, 'VideoResolution');
imWidth = vidRes(1);
imHeight = vidRes(2);
nBands = get(vid, 'NumberOfBands');
hImage = image( zeros(imHeight, imWidth, nBands) );
figSize = get(hFig,'Position');
figWidth = figSize(3);
figHeight = figSize(4);
set(gca,'unit','pixels','position',[ 275 170 250 250 ]);
%'position',[ ((figWidth - imWidth)/2) ((figHeight - imHeight)/2) imWidth imHeight ]);
preview(vid,hImage);
wait(5000);
function GetFR_Callback (src, evt)
index= get(src,'Value');
set(msgs,'String',strcat('Framerate = ',num2str(framerate), ' fps'));
end
function startR_Callback(src, evt)
set(msgs,'String','RECORDING VIDEO...');
set(vid,'FramesPerTrigger',100);
interval = 1;
set(vid,'FrameGrabInterval',interval);
capturetime = 20;
numframes = floor(capturetime * framerate / interval);
set(vid,'FramesPerTrigger', numframes);
set(vid,'LoggingMode','disk');
fname = strcat('I',int2str(cnt),'.avi');
avifil = avifile(fname,'fps',framerate,'colormap','rgb');
set(vid,'DiskLogger',avifil);
start(vid);
wait(vid,Inf);
avifil = get(vid,'DiskLogger');
avifil = close(avifil);
set(msgs,'String','FILE RECORDED!!!');
% stoppreview(vid);
cnt = cnt+1;
% delete(vid);
% clear vid;
return;
end
function closeW_Callback(src, evt)
stoppreview(vid);
delete(vid);
close(gcf);
close all;
clear all;
end
function Compress_Callback(src, evt)
set(msgs,'String','COMPRESSING.');
fname = strcat('I',int2str(ccnt),'.avi');
cname = strcat('C',int2str(ccnt),'.avi');
dctavi = avifile(cname,'colormap','rgb');
finfo = aviinfo(fname);
icmp = finfo.FileSize;
for i = 1:finfo.NumFrames
mov = aviread(fname ,i);
[a,map] = frame2im(mov); %video converted 2 image
coeff=1000;
red = double(a(:,:,1));
green = double(a(:,:,2));
blue = double(a(:,:,3));
red_dct=dct2(red);
green_dct=dct2(green);
blue_dct=dct2(blue);
red_pow = red_dct.^2;
green_pow = green_dct.^2;
red_pow=red_pow(:);
green_pow=green_pow(:);
blue_pow=blue_pow(:);
[B_r,index_r]=sort(red_pow);

JERS/Vol. II/ Issue III/July-September, 2011/147-153

 Journal of Engineering Research and Studies E-ISSN0976-7916

[B_g,index_g]=sort(green_pow);
[B_b,index_b]=sort(blue_pow);
index_r=flipud(index_r);
index_g=flipud(index_g);
im_dct_r=zeros(size(red));
im_dct_g=zeros(size(green));
im_dct_b=zeros(size(blue));
im1=zeros(size(red,1),size(red,2),3);
for ii=1:coeff
im_dct_r(index_r(ii))=red_dct(index_r(ii));
im_dct_g(index_g(ii))=green_dct(index_g(ii));
im_dct_b(index_b(ii))=blue_dct(index_b(ii));
end
im1(:,:,1)=im_dct_r;
im1(:,:,2)=im_dct_g;
im1(:,:,3)=im_dct_b;
im1=uint8(im1);
dctavi = addframe(dctavi,im1) ; % dct avi file
end
ccnt = ccnt +1;
dctavi = close(dctavi);
% close all
% clear all
cinfo = aviinfo(cname);
ccmp = cinfo.FileSize;
cmpr = double ((ccmp / icmp) * 100);
set(msgs,'String','FILE COMPRESSED!!!');
set(cr,'String',num2str(cmpr));

Fig 10. DCT applied to a video frame

Original image Leveled off image

After 1-D DCT Total DCT

Fig 11. Total DCT of a video frame

JERS/Vol. II/ Issue III/July-September, 2011/147-153

 Journal of Engineering Research and Studies
Fig 12. Rate / Distortion curves with & without ABT
Fig 13. Motion compensated frame
6.0 CONCLUSION:
Digital video compression, although only recently
becoming a standardized technology, is strongly
based upon the information coding technologies
developed over the past 40 years. The large variety of
bandwidth and video quality requirements for the
transmission and storage of digital video information
has demanded that a variety of video compression
techniques and standards be developed. The major
international standards recommended by IS0 and the
ITU make use of common video coding methods.
They address avast landscape of application
requirements, from low- to high bitrate environments,
as well as stored video and multimedia to real-time
videoconferencing and high-quality broadcast
television.
The near future will drive video compression
systems to incorporate support for more interactive
functions, with the ability to define and download
new functions to the encoder. New encoding methods
currently being explored by the MPEG-4 and MPEG-
7 standards look toward object-based encoding in
which the encoder is not required to follow the
international video transmission signal formats.
These object-based techniques are expected to
produce significant improvements in both encoder
efficiency and functionality for the end user.
JERS/Vol. II/ Issue III/July-September, 2011/147-153
E-ISSN0976-7916
REFERENCES
1. Ken Cabeen and Peter Gent, Image Compression &
Discrete Cosine Transform
2. Alexandre Krivolulhefts, A Method for Progressive
Near Lossles Compression, Proceedings of IEEE
International Conference on Image Processing, pp 185-
188, 2003.
3. I.Avcibas, N.Memon, B.Sankur and K. Sayood, A
Progressive Lossless/ Near Loss Less Image
Compression Algorithm, IEEE signal processing
letters 9(10), pp 312-314, 2002.
4. W.B. Pennebaker, J.L. Mitchell, "The JPEG Still
Image Data Compression Standard", Van Nostrand
Reinhold, 1993.
5. V Bhaskaran and K. Konstantinides, "Image and Video
Compression Standards: Algorithms and
Architectures", 2nd ed. Kluwer Academic Publishers,
1997.)
6. Nathenial J, August , Dong Sam Ha, Low Power
Design of DCT and IDCT for Low Bit rate Video
Codecs, IEEE Transactions on Multimedia Vol.6, No.
3, June 2004,pp(414 - 420).
7. Jay Loomis and Mike Wasson, VC-1 Technical
Overview, Microsoft Corporation, Oct. 2007.
8. Andrew B. Watson, Image Compression using
Discrete Cosine Transform, Mathematica Journal,
4(1), 2004, pp 84-88.
9. M.A Robertson and R.L. Stevenson, DCT
Quantization Noise in Compressed Images, IEEE
Transactions on Circuits & Systems for Video
Technology, Vol. 15, No.1 , Jan. 2005, pp 27 - 38.
10. S. Dhahri, A. Zitouni, H. Chaouch, and R. Tourki,
Adaptive Motion Estimator Based on Variable Block
Size Scheme, World Academy of Science, Engineering
& Technology, 50, 2009, pp 384 390.