Nghiên cứu giải pháp nâng cao tốc độ xử lý tín hiệu trong hệ thống thị giác nổi tt tiếng anh

MINISTRY OF EDUCATION AND TRAINING

MINISTRY OF NATIONAL DEFENSE

ACADEMY OF MILITARY SCIENCE AND TECHNOLOGY

DOAN VAN TUAN

RESEARCH ON THE APPROACH TO IMPROVE SIGNAL
PROCESSING SPEED IN THE STEREO VISION SYSTEM

Specialization: Electronic Engineering
Code: 9 52 02 03

SUMMARY OF PhD THESIS IN ELECTRONIC ENGINEERING

Ha noi, 2019


The thesis has been completed at:
ACADEMY OF MILITARY SCIENCE AND TECHNOLOGY

Scientific supervisors:
1. Dr. Ha Huu Huy

2. Assoc. Prof. Dr. Bui Trung Thanh
Reviewer 1: Assoc. Prof. Dr. Hoang Manh Thang
Hanoi University of Science and Technology
Reviewer 2: Assoc. Prof. Dr. Le Nhat Thang
Posts and Telecommunications Institute of Technology
Reviewer 3: Dr. Vu Le Ha

Academy of Military Science and Technology

The thesis was defended at the Doctoral Evaluating Council at
Academy level held at Academy of Military Science and Technology at
……..., date 2019.

The thesis can be found at:
- The library of Academy of Military Science and Technology.
- Vietnam National Library.


INTRODUCTION
1. The necessity of the thesis
Today, science and technology has developed strongly, especially
since the industrial revolution 4.0 was initiated from Germany in 2013. One
of the factors that dominated the industrial revolution 4.0 is that robots will
gradually replace labors working in factories. Therefore, the robot must
process information in a three-dimensional environment (3D) through the
vision system to orient, locate, identify and accurately locate the
surrounding objects called stereo vision or 3D robot vision. In addition,
stereo vision is also applied in identification, regeneration, positioning,
surgery, self-propelled vehicles, mapping and in art.
People want to create a robot vision system like human vision, the
simplest vision system is to use stereo camera including two cameras
combined with embedded processing system to replace two human eyes.
Stereo camera information is processed through algorithms based on
processing devices such as CPU, DSP, GPU, FPGA and ASIC combined
with implementation languages such as Matlab, OpenCV, CUDA .... Such a
system is called a stereo vision system. The major challenges for the wellknown stereo visio system are the stereo camera as the data source from the
stereo camera image is increasing, the execution speed requires real-time

response, high reliability and finite memory capacity. To solve this
problem, one of the most effective solutions is to develop processing
algorithms, while the processing platforms have not yet developed
according to human needs.
2. The objectives of the thesis
Study specific approachs to improve belief propagation algorithms to
speed-up of execution and reduce the amount of memory required when
executing the disparity map of dense two-frame high resolution stereo
camera in the stereo vision system application for 3D robot vision.
3. Research objectives and scope of the thesis
-The thesis focuses on studying the solution to reduce the energy


function of the belief propagation algorithms implements the disparity map
of dense two-frame high resolution stereo camera image in the stereo vision
system.
- Stereo camera images are taken from the test data set.
- Study and propose solutions to improve speed-up BP algorithm to
advance the effectiveness of the disparity map implementation.
4. Research methodology of the thesis
The thesis focuses on researching solutions to optimization energy
function of the belief propagation algorithms implements the disparity map
of dense two-frame high resolution stereo camera. Analyze improvement
belief propation algorithms and propose solutions to reduce the the energy
function of the belief propagation algorithm and select appropriate
processing platform to achieve the purpose of the thesis. From
mathematical analysis, parameterization of the parameters, the thesis uses
simulation tools, data from the test data set to prove the correctness of the
research results.
5. The scientific and practical contributions of the thesis

The disparity map of stereo camera has a very important role in 3D
Robot vision. From the disparity map, combined with triangulation, the
depth map and the distance from the camera to the object can be estimated.
This technique is widely applied in industry, robotics, surgery, selfpropelled vehicles, localization and mapping.
The thesis has proposed two solutions to reduce the cost function for
belief propagation algorithm. The first solution is to reduce the number of
nodes in the Markov random field model through loops using the CTF
method to level 1. The second solution is to combine the local census
transform algorithm and global belief propagation algorithm has improved
the cost reduction of the initial start button when implementing the belief
propagation algorithm's message update.
6. Research content and structure of the thesis
The whole thesis consists of 137 pages presented in 3 chapters, 40
drawings, 29 tables and 14 charts.


Chapter 1:
OVERVIEW OF STEREO VISION AND SIGNAL PROCESSING IN
THE STEREO VISION SYSTEM
1.1. Overview of stereo vision
Stereo vision is a very important component in computer vision and
has been researched and developed by many scientists in the last two
decades. [46]. Stereo vision system is widely applied in many areas such as
robots, self-propelled vehicles, medical, arts, entertainment and especially
in industrial networks 4.0 [59]. People want to create a robot vision system
that works in a 3D environment similar to human vision called stereo vision
system like Figure 1.1, when robots and humans work together, interactive
[107]
.
Image information

Image processing

Application

Figure 1.1. Scheme block of stereo vision system
1.2. Model camera
1.3 Calibration methods
The camera calibration method will determine the speed and reliability
of the camera's internal and external parameters. Currently there are a
number of classic image calibration methods such as Hall [39], Salvi [37],
Tsai [91] and Weng [76] based on the corresponding camera models. Each
model will have appropriate calibration methods and have different
advantages and disadvantages.
1.4. Rectification methods
Rectification methods optimize finding homologous points in the
stereo camera images and improve image processing reliability. The
rectification method is divided into two types. In the first form, the
rectification methods after calibration [9], [105]. The second form, the
rectification methods performed without calibration [26].


1.5. Stereo matching algorithms
Over the past two decades, many matching algorithms have been
proposed [46]. The matching algorithm is classified according to the stereo
camera image. Matching algorithms for sparse two-frame high resolution
stereo camera images such as SIFT [10], SURF [66] are often used for
stereo vision systems that require high speed and memory capacity low
requirements however do not require high reliability, often applied to
navigation systems, mapping or SLAM [36] and self-propelled vehicles.

Matching algorithms for dense two-frame high resolution stereo camera
images as [7], [44] which are often used for stereo vision systems that
require high reliability, often applied to industrial product inspection
systems, 3D visual system of robot vision and in surgery or object
reproduction, however, large computational complexity and high memory
requirements. Matching algorithms for dense two-frame high resolution
stereo camera images have three main types: local algorithms [15], [101],
global algorithms [48], [78] and semi-global algorithm [24], [90].
1.6. Hardware processing method in stereo vision
- Method CPU
- Method DSP
- Method GPU
- Method FPGA/ASIC
1.7. Evaluation hardware processing method in stereo vision
From CPU → DSP → GPU → FPGA → ASIC, processing efficiency
increases sequentially, while costs and power consumption decrease
accordingly. Stereo matching algorithms have more flexibility and short
development cycles, while hardware performs a longer design cycle with
less design flexibility because at the same time the algorithm must be
considered optimally and collect hardware map. From a practical point of
view, the stereo vision processing hardware system needs to be more
accessible to real-time stereo vision systems because it consumes low
power and is cheaper.


1.8. Research directions to improve the efficiency of the stereo vision system
- Image segmentation or hierarchy optimization
- Occlusion and consistency handling
- Matching cost & energy optimization improvement
- Cooperative optimization

- Efficient memory arrangement method
- Advanced VLSI design method
1.9. Conclusion of chapter 1
Chapter 1 presents an overview of the main components of the stereo
vision system consisting of two main blocks of image information block
and image processing block. Each component has also been analyzed and
given an assessment of its role in the system. The image information block
consists of two main components: stereo camera and camera calibration.
This block provides stereo camera parameters such as image size and depth
disparity, internal parameters and parameters outside the stereo camera.
The parameters also affect the reliability of the system. The image
processing unit will determine the effectiveness of the system, including
software and hardware. Software is programming languages that perform
processing algorithms including image rectification algorithms, matching
algorithms. In particular, the role of matching algorithms will primarily
affect the efficiency of the system. Hardware is the processing platform for
implementing software solutions and it also plays a role in improving the
efficiency of the stereo vision system. In addition, the match choice
between the processing platform and the matching algorithm also
influences the performance of the stereo vision system.
The selected hardware is the GPU processing platform of Nvidia GXT
750Ti with 2GB, 460 memory and 128 bit bandwidth using CUDA 7.5
software and QT creator compiler combined with Intel core i7 CPU, RAM
8 GB with Windows 8.1 operating system. The GPU processing platform is
selected because it supports parallel processing structures, has multiple
processor cores, broadband and memory is increasingly being increased in
accordance with the experimental program of the thesis.


Chapter 2:

RESEARCH BELIEF PROPAGATION ALGORITHMS AND
BUILD THE METHODS TO IMPROVE SIGNAL PROCESSING
SPEED IN THE STEREO VISION SYSTEM
2.1. Markov random field
Markov random field is a branch of probability theory. Markov
random field is used as a tool to processing image data modeling, combined
with winner-take-all algorithms. In addition, the Markov random field is
used as a means of generating inference results on images. The inferences
related to basic image and frame structure will solve problems such as
image reconstruction, image segmentation, stereo vision and creating object
labeling. The markov random field model usually has two forms: grid-like
structure and part-based structure.
2.2. Belief propagatin
Belief propagation uses messages containing the disparity values of
the corresponding points and moves between nodes according to iterative
methods to perform inference on the graph model. This method provides
accurate inferences with part-like structure models and provides
approximate reasoning for grid-like structure. A belief propagation algorithm
is used to identify maximum a posteriori (MAP) in markov random field
models for stereo vision problems.
2.3. Census transform
Census transform is a non-parametric transformation algorithm, it does
not depend on the light conditions of the image [86]. The operating
principle of census transform is to convert each pixel into a bit length bit
string with local space architecture. For each neighboring pixel except the
center point will transform respectively into a bit in the sequence of N bits
according to threshold if the value of intensity is close, the neighboring bit
is greater than the central bit strength value corresponding to a bit equal to
1, then the bit is 0.
2.4. Approachs to improve processing speed-up of belief propagation

algorithm


- Parallel calculations.
- Reduce computational complexity.
- Reduce the amount of memory required when performing.
- Minimum update messages.
- Optimize the way to access memory.
- Reuse memory.
- Improve reliability.
- Cooperative optimization.
- Select appropriate processing algorithms and handling platform
The thesis proposes two solutions to improve the processing speed-up
for the belief propagation algorithm is a cost reduct energy function and a
combined optimization solution.
2.5 Propose solutions to reduce cost functions
2.5.1. Proposed algorithm 1
The model of proposed algorithm coarse to fine belief propagation
(CFBP: proposed algorithm 1) is built based on the markov random field
model in grid-like structure, node with 4 neighborhood as Figure 2.16.
Consider G = (E, V) where G is a graphical model, E is a set of nodes, V is
a set of edges. The node is the label that is assigned the value of the
intencity disparity of the stereo corresponding point of the stereo camera,
often called data value or data function. The edge is the label assigned the
disparity value of the two neighboring labels, often called cost smooth or
smooth function.

Figure 2.16. Scheme of proposed algorithm model 1

From the model proposed algorithm 1 shows that the proposed
algorithm used the coarse-to-fine (CTF) level 1 method as shown in Figure
2.17 to reduce the number of nodes after loops. Method CTF is used to
deduce the reduction of the number of nodes by levels. After executing
CTF level l, the number of nodes on the current loop will decrease S = 2x2l
times the number of nodes in the previous loop. The cost value for doing 4node reasoning on a node is determined by the formula (2.36). The message
in the proposed algorithm passing in a parallel scheme as shown in Figure
2.18. The initial start node selected is the node labeled (0, 0) with the initial
'
 0.
message values set to be m0'  0 and m0,0

Figure 2.17. Structure scheme CTF
level 1

Figure 2.18. Scheme passing
message
The energy function at CTF level 1 is given by
1
E CTF 1 ( x)    ( xi )
(2.36)
4 i[1,4]
The number of iterations performed CTF level 1 is k 2' and is

determined by the formula (2.37). Considering the stereo camera image
resolution is m, n and k where m is the number of pixels in the row, n is the
number of pixels in the column and k is the number of depth diaparity of
the image.
k2'   log 2 m
(2.37)

Number of loops k1' be done in every CTF level 1 is defined as:
k
k1' 
log2 m  1
+ The energy function is given by

(2.38)


E ' ( x)   ' ( xi ) 
iE

  (x , x )
'

i

i , jV

(2.39)

j

The discontinuity cost  ' ( xi , x j ) is generally based on the difference
between labels, the labels correspond to the possible disparities, and the
cost of assigning a pair of labels to neighboring pixels is based on the
degree of difference between the labels.
(2.40)
 ' ( xi , x j )   ' ( xi  x j )
E ' ( x)   ' ( xi ) 

iE



i , jV

 ' ( xi  x j )

(2.41)

The proposed algorithm 1 works by passing messages around the graph
defined by the four-connected image grid. The method is iterative, with
messages from all nodes being passed in parallel. Each message is a vector
of dimension given by the number of possible labels, k ' . Let mi' j be the
message that node i sends to a neighboring node j at iteration t. When using
mi'0 j  0 and at each iteration new messages are computed in the following
way,

mi'(t ) j ( x j )  min(  ' ( xi  x j )   ' ( xi ) 
xi



sN ( i )\ j

t 1)
ms'(
i ( xi ))

mi'(t ) j ( x j )  min(  ' ( xi  x j )  h' ( xi ))

then h' ( xi )   ' ( xi ) 



xi

sN ( i )\ j

(2.42)
(2.43)

t 1)
ms'(
i ( xi )

The smooth cost function is determined by linear model.
 ' ( xi  x j )  min( p' xi  x j , q' )

(2.44)

Messages update are computed in the following way,
mi'(t ) j ( x j )  min(min( p' xi  x j , q' )  h' ( xi ))

(2.45)

After T iterations a belief vector is computed for each node:
b' j ( x j )   ' ( x j )   mi'  j ( x j )

(2.46)

xi

iN ( j )

Node x '*j is selected and determined by the formula:
x'*j  arg min b' j ( x j )

(2.47)


The energy function is given by:
E ' ( x)   I L ( x, y)  I R ( x  xi , y) 
iE



i , jV

min( p ' xi  x j , q ' )

(2.48)

where I L ( x, y ) is the gray level of the left image at the coordinates ( x, y ) ,
I R ( x  xi , y) is the gray level of the right image at the coordinates
( x  xi , y ) of stereo camera image.
2.5.2. Proposed algorithm 2
The model of proposed algorithm coarse to fine change space belief
propagation (CFCSBP: proposed algorithm 2) has the same structure as the
proposed algorithm 1 model as Figure 2.20, however, there is a diffrence
between these two models is that while the proposed algorithm 1 must

perform the number of loops equal to the number of disparity of the image
then the proposed algorithm 2 has a number of loops that vary according to
the Z '' coefficient by formula (2.50) compared to the depth disparity of the
image.

Figure 2.20. Scheme of proposed algorithm 2 model
Considering the stereo camera image resolution is m, n and k, where m
is the number of pixels in the row, n is the number of pixels in the column
and k is the number of depth disparity of the image. The number of coarse
to fine CTF level 1 is determined by formula (2.49) for reasons such as
selection k 2' .
k2''   log 2 m

''
1

Number of loops k be done in every CTF level 1 is defined as:

(2.49)


k1'' 

k
Z (log 2 m  1)
''

(2.50)

where Z '' is the coefficient of depth change.

Calculating the cost value for passing the message of the proposed
algorithm 2 is the same as that of the proposed algorithm 1 except that the
proposed algorithm 1 must perform k '  k1'  k2' the loop while the proposed
algorithm 2 performs the k ''  k1''  k2'' the loop
2.6. Propose solutions to cooperative optimization
2.6.1. Proposed algorithm 3
The model of proposed algorithm census transform belief propagation
(CTBP: proposed algorithm 3) is built based on the markov random field
model in grid-like structure, node with 4 neighborhood as Figure 2.22.
Consider G = (E, V) where G is a graphical model, E is a set of nodes, V is
a set of edges. The node is the label that is assigned the value of the
intencity disparity of the stereo corresponding point of the stereo camera,
often called data value or data function. The edge is the label assigned the
disparity value of the two neighboring labels, often called cost smooth or
smooth function. Let V1, V2, V3, V4 and E1, E2, E3, E4 respectively nodes and
edges of part 1, part 2, part 3 and part 4 of the proposed algorithm 3 model.

Figure 2.22. Scheme of proposed algorithm 3 model
From the proposed algorithm 3 model show that, the start nodes for
passing message to be labeled (0, 0) on belief propagation algorithm model


m n
has been replaced by node which is the labeled node  ,  on the
 2 2
proposed algorithm 3 model. The process of finding the corresponding
m n
point for the node is labeled  ,  is done via census transform method
 2 2

as formula (2.35) with window 3x3. The image is rectified as 4 parts
through the corresponding point determined by CT method. Each part will
be passing message according to BP algorithm with the pass scheme as
m n
shown in Figure 2.18, the start node is the node labeled  ,  . All 4
 2 2
parts of the model will be implemented simultaneously.
+ Calculate the cost value
The 4 parts of the model are the same size, the method of calculating
the cost value is the same.
+ Calculate the cost value for 1 part of model
Calculating the cost value of passing the message of proposed
algorithm 3 is the same calculating the cost value of passing the message of
the proposed algorithm 1, only different in the size of the propagation
nodes. The proposed algorithm 1 has a start pass node that is labeled (0,0)
while the proposed algorithm 3 has a start pass node that is labeled
m n
 , .
 2 2
2.6.2. Proposed algorithm 4
Model of proposed algorithm census transform change space belief
propagation (CTCSBP: proposed algorithm 4) is built based on the MRF
model in grid-like structure, node with 4 neighborhood the same proposed
algorithm 3 as Figure 2.24. However, there is another difference between
these two models is that while proposed algorithm 3 must perform the
number of loops equal to the number of disparity of the image, proposed
algorithm 4 has the number of loops changed according to the ratio of Z ''''
according to the formula (2.71 ) compared to the node disparity. Let V1, V2,

V3, V4 and E1, E2, E3, E4 respectively nodes and edges of part 1, part 2, part
3 and part 4 of the proposed algorithm 4 model.
Calculating the cost value for passing the message of the proposed
algorithm 4 is the same as that of the proposed algorithm 3 except that the
proposed algorithm 3 must perform k '''  k the loop while the proposed
algorithm 4 performs the k '''' the loop.

Figure 2.24. Scheme of proposed algorithm 4 model.
k
k ''''  ''''
(2.71)
Z
2.7. Evaluation methodology
Two general approaches to this are to compute error statistics with
respect to some ground truth data and to evaluate the synthetic images
obtained by warping the reference or unseen images by the computed
disparity map [29]:
+RMSE (root-mean-squared-error) measured in disparity units between
the computed disparity map dC(x,y) and the ground truth map dT(x,y):
1

2
1
2
R    dc ( x, y )  dT ( x, y ) 
 N ( x, y )

+ Percentage of bad matching pixels:
1
B   ( d C ( x, y )  d T ( x , y )   d )

N ( x, y )

(2.82)

(2.83)


2.8. Conclusion of chapter 2
Chapter 2 of the thesis presents the theoretical basis of the belief
propagation algorithm is a markov random field including graph theory
combined with probability theory. Research and application of belief
propagation algorithm to determine the disparity map of dense two-frame
high resolution stereo camera image. Analysis and evaluation of belief
propagation algorithms improvements implemente disparity map of stereo
cameras, there by giving directions to improve and improve signal
processing speed of belief propagation algorithm application in stereo
vision system.
From the analysis and evaluation of the algorithms implemented, the
thesis has proposed two solutions to improve the signal processing speed of
the belief propagation algorithm application for stereo visison is reduce
cost function solution and cooperative optimization solution. Both solutions
have a model based on the markov random field in grid-like structure, node
with 4 neighborhood. The reduce cost function solution is represented by
two proposed algorithms that are coarse to fine belief propagation (CFBP:
proposed algorithm 1) and coarse to fine change space belief propagation
(CFCSBP: proposed algorithm 2). The cooperative optimization solution is
a combination of local CT algorithms and global algorithms BP,
represented by two proposed algorithms that are census transform belief
propagation (CTBP: proposed algorithm 3) and census transform change
space belief propagation (CTCSBP: proposed algorithm 4).

Chapter 3
EXPERIMENT AND EVALUATION OF THE RESULTS
3.1. Tools and experimental data
test
The experimental system as
shown in Figure 3.1 with the PC
configuration described in Table 3.1
and stereo camera image in the test
data set [30] is described in Table 3.2.
Figure 3.1. Experimental
system


Table 3.1. Describe the PC Desktop configuration
Hardware
Software
CPU RAM
Graphic card
Operating
Application
system
software
Intel 8GB Geforce GTX750 Ti
Window
QT Creator 5.8
core
RAM: 2GB
8.1
OpenCV 3.0
i7

Core: 460
64 bit
Visual Studio 2013
BUS: 128 bit
CUDA
Table 3.2. Test data set
Image
Right
Disparity
Symbol
Size
Disparity Left image
name
image
map true
#1

Baby

620x555

300

#2

Aloe

641x555

270

#3

Cloth

626x555

290

#4

Flower
656x555
pots

251

#5

Bowling

665x555

240

#6

Book

695x555

200

3.2.Root mean squared error (RMSE)
3.3. Experiments and results
3.3.1. Belief propagation standard


The processing speed of implementing the disparity map of belief
propagation algorithm [78] in the system Figure 3.1 has the parameters
described as Table 3.1 and test data Table 3.2 is shown in Table 3.3.
Table 3.3. Processing speed of BP algorithm (ms)
Image
#1
#2
#3
#4
#5
#6
TT
BP
439
457
442
473
478
494
3.3.2. Proposed algorithm 1
Disparity map results of camera stereo image Table 3.2 when using the
proposed algorithm 1 is shown in Figure 3.3.

(a)
(b)
(c)
(d)
(e)
(f)
Figure 3.3. Disparity map uses proposed algorithm 1: (a), (b), (c), (d), (e)
,(f) is the corresponding disparity map of the images #1, #2, #3, #4, #5,#6.
The processing speed of implementing the disparity map of proposed
CFBP algorithm in the system Figure 3.1 has the parameters described as
Table 3.1 and test data Table 3.2 is shown in Table 3.6.
Table 3.6. Processing speed of proposed CFBP algorithm (ms)
Image
#1
#2
#3
#4
#5
#6
TT
CFBP
206
217
213
224
227
235
3.3.3. Proposed algorithm 2
Disparity map results of camera stereo image Table 3.2 when using the

proposed algorithm 2 with the coefficient Z ''  3 , is shown in Figure 3.4.

(a)
(b)
(c)
(d)
(e)
(f)
Figure 3.4. Disparity map uses proposed algorithm 2: (a), (b), (c), (d), (e),
(f) is the corresponding disparity map of the images #1, #2, #3, #4, #5, #6.


The processing speed of implementing the disparity map of proposed
''
algorithm 2 with the coefficient Z  3 in the system Figure 3.1 has the
parameters described as Table 3.1 and test data Table 3.2 is shown in Table 3.9.

Table 3.9. Processing speed of proposed algorithm 2 (ms)
#1
#2
#3
#4
#5
#6
TT
CFCSBP
191
199
195
203

204
211
3.3.2. Proposed algorithm 3
Disparity map results of camera stereo image Table 3.2 when using the
proposed algorithm 3 combined with CT algorithm with 3x3 window and
length scan xCT  10 , is shown in Figure 3.5..
Image

(a)
(b)
(c)
(d)
(e)
(f)
Figure 3.5. Disparity map uses proposed algorithm 3: (a), (b), (c), (d), (e) ,
(f) is the corresponding disparity map of the images #1, #2, #3, #4, #5, #6.
The processing speed of implementing the disparity map of proposed
CTBP algorithm in the system Figure 3.1 has the parameters described as
Table 3.1 and test data Table 3.2 is shown in Table 3.12.
Table 3.12. Processing speed of proposed CTBP algorithm (ms)
Image
#1
#2
#3
#4
#5
#6
TT
CTBP
182

185
182
187
188
191
3.3.2. Proposed algorithm 4
Disparity map results of camera stereo image Table 3.2 when using the
proposed algorithm 4 with the coefficient Z ''''  3 , is shown in Figure 3.6.

(a)
(b)
(c)
(d)
(e)
(f)
Figure 3.6. Disparity map uses proposed algorithm 4: (a), (b), (c), (d), (e),
(f) is the corresponding disparity map of the images #1, #2, #3, #4, #5, #6.


The processing speed of implementing the disparity map of proposed
''''
algorithm 4 with the coefficient Z  3 in the system Figure 3.1 has the
parameters described as Table 3.1 and test data Table 3.2 is shown in Table 3.15.

Table 3.15. Processing speed of proposed algorithm 4 (ms)
#1
#2
#3
#4
#5

#6
TT
CTCSBP
83
84
83
89
92
94
3.4. Evaluation proposed algorithm
3.4.1. Evaluation proposed algorithm 1 compare with BP algorithm
The processing speed of the disparity map execution of the stereo
camera depends on factors such as the execution algorithm (programming
skills), the compiler, the configuration of the system and the input data.
Therefore, in order to evaluate the performance processing speed of the
proposed algorithm 1 and BP algorithm [78], the thesis has implemented
these two algorithms on the same experimental system as Figure 3.1 with
the parameters shown in Table 3.1 and test data set as shown in Table 3.2.
The results of the processing speed comparison between the two algorithms
are described in Table 3.18 and Diagram 4.1.
Table 3.18. Compare the processing speed of proposed 1 and BP algorithms (ms)
Image
BP
Proposed 1
Speedup
#1
439
206
113,11%
#2

457
217
110,60%
#3
442
213
107,51%
#4
473
224
111,16%
#5
478
227
110,57%
#6
494
235
110,21%
Table 3.18 shows that, for test images of equal size and different
disparitys, the execution time is almost unchanged when performing the
same algorithm. This shows that the processing speed of the disparity map
implementation does not depend on the complexity and difference of the
sample image, but depends on the resolution of the image. In addition,
Table 3.18 also shows that the processing speed of the proposed algorithm
1 increased 113.11% compared to the standard BP algorithm when
Image


implementing image # 1 in Table 3.2. To be intuitive, the thesis describes

comparing the performance rates of the two algorithms through diagram 4.1.
Diagram 4.1. Compare the processing speed of CFBP and BP algorithms (ms)

3.4.2. Evaluation proposed algorithm 2 compare with BP algorithm
In order to evaluate the processing speed of the proposed algorithm 2,
thesis compares the processing speed of the proposed algorithm with the
processing speed of BP algorithm implementation on the same
experimental system as Figure 3.1 with the parameters shown in Table 3.1,
the test data set as shown in Table 3.2 and the coefficient depth changes
Z ''  3 . The processing speed performance results between the proposed
algorithm 2 and the BP algorithm are described in Table 3.21.
Table 3.21 shows that, for test images of equal size and different
disparitys, the execution time is almost unchanged when performing the
same algorithm. This shows that the processing speed of the disparity map
is not dependent on the complexity and the disparity of the sample image,
but depends on the resolution of the image.
Table 3.21. Compare the processing speed of proposed 2 and BP algorithms (ms)

Image
#1
#2
#3
#4
#5
#6

BP
439
457
442

473
478
494

Proposed 2
191
199
195
203
204
211

Speedup
129,84%
129,65%
126,67%
133,00%
134,31%
134,12%


In addition, to be intuitive, the thesis describes comparisons,
evaluation the processing speed of the proposed algorithm 2 and BP
algorithm through Diagram 4.4.

Diagram 4.4. Compare the processing speed of CFCSBP and BP algorithms (ms)

3.4.3. Evaluation proposed algorithm 3 compare with BP algorithm
The processing speed of the disparity map execution of the stereo
camera depends on factors such as execution algorithm, compiler, system

configuration and input data. Therefore, to evaluate the performance
processing speed of the proposed algorithm 3 and BP algorithm [78], the
thesis has implemented these two algorithms on the same experimental
system as Figure 3.1 with the parameters shown. in Table 3.1 and test data
set as shown in Table 3.2. The results of processing speed comparison
between two algorithms are described in Table 3.24 .
Table 3.24. Compare the processing speed of proposed 3 and BP algorithms (ms)

Image
BP
Proposed 3
Speedup
#1
439
182
141,21%
#2
457
185
147,03%
#3
442
182
142,86%
#4
473
187
152,94%
#5
478

188
154,26%
#6
494
191
158,64%
Table 3.24 shows that, for test images of equal size and different disparitys,
the execution time is almost unchanged when performing the same
algorithm. This shows that the processing speed at which the disparity map


is not depend on the complexity and disparity of the image, but depends on
the resolution of the image. In addition, Table 3.24 also shows the
processing speed of the proposed algorithm 3 to increase 141.21%
compared to the standard BP algorithm when implementing image # 1 in
Table 3.2. To be intuitive, the thesis describes comparing the performance
rates of the two algorithms through diagram 4.7.

Diagram 4.7. Compare the processing speed of CTBP and BP algorithms (ms)
3.4.4. Evaluation proposed algorithm 4 compare with BP algorithm
In order to evaluation the processing speed of the proposed algorithm
4, the thesis compares the processing speed of the proposed algorithm with
the processing speed of BP algorithm implementation on the same
experimental system as Figure 3.1 with the parameters shown in Table 3.1,
test data set as shown in Table 3.2 and the coefficient depth changes
Z ''''  3 . The performance results between the proposed algorithm and BP
algorithm are described in Table 3.26.
Table 3.26. Compare the processing speed of proposed 4 and BP algorithms (ms)

Image

#1
#2
#3
#4
#5
#6

BP
439
457
442
473
478
494

Proposed 4
83
84
83
89
92
94

Speedup
428,92%
444,05%
432,53%
431,46%
419,57%
425,53%

To be intuitive, the thesis describes comparing the performance rates of
the two algorithms through diagram 4.9.

Diagram 4.9. Compare the processing speed of CTCSBP and BP algorithms (ms)

3.4.5. Overall comparison proposed algorithm
To compare the proposed algorithms, the thesis employs the proposed
algorithms on the same experimental system with the same compiler and
input data. Focusing on comparing performance processing speed, in
addition, the thesis also considers factors such as memory capacity
requirements and reliability. Table 3.29 shows the processing speed at
which the proposed algorithms are implemented when implemented on the
same system as Figure 3.1 with the configuration as shown in Table 3.1 and
input data as shown in Table 3.2 with the choice coefficient depth change
Z ''  3 and coefficient depth change Z ''''  3 .
Table 3.29. Compare the processing speed of proposed algorithms (ms)
Image
BP
Proposed 1 Proposed 2 Proposed 3 Proposed 4
#1
439
206
191
182
83
#2
457
217

199
185
84
#3
442
213
195
182
83
#4
473
224
203
187
89
#5
478
227
204
188
92
#6
494
235
211
191
94
Table 3.29 shows that the proposed algorithms have improved
performance compared to standard BP algorithms. In the proposed

algorithms, the proposed algorithm 2 with the coefficient depth change
Z ''''  3 for the best processing speed. To be intuitive, the thesis represents
comparing the performance processing speed of proposed algorithms
through diagram 4.12.

Diagram 4.12. Compare the processing speed of proposed algorithms (ms)
3.5. Conclusion of chapter 3
Chapter 3 has experimented with proposed algorithms and belief
propagation standard algorithms on the same platform with the same
compiler and input test data set. Three main factors to compare and
evaluate algorithms are performance speed, required memory capacity and
reliability. The thesis has evaluated each proposed algorithm with belief
propagation standard algorithm according to three factors: performance
processing speed, required memory capacity and reliability. However, in
the above three factors, the thesis focuses on evaluating the processing
speed of implementation is mainly. Most of the proposed algorithms have
improved the processing speed of execution, the amount of memory
required but must pay for reliability.
In the proposed algorithms, the proposed algorithm 4 has the most
advantages, quickly improving the execution processing speed while the
reliability is negligible.