/
Libro.lyx
3100 lines (2328 loc) · 81.4 KB
/
Libro.lyx
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
#LyX 2.0 created this file. For more info see http://www.lyx.org/
\lyxformat 413
\begin_document
\begin_header
\textclass report
\begin_preamble
\usepackage[usenames,dvipsnames]{xcolor}
\usepackage[colorlinks, linkcolor=black, citecolor = Sepia]{hyperref}
\end_preamble
\options openright
\use_default_options true
\maintain_unincluded_children false
\language english
\language_package default
\inputencoding utf8x
\fontencoding global
\font_roman default
\font_sans default
\font_typewriter default
\font_default_family default
\use_non_tex_fonts false
\font_sc false
\font_osf false
\font_sf_scale 100
\font_tt_scale 100
\graphics default
\default_output_format default
\output_sync 0
\bibtex_command bibtex8
\index_command default
\paperfontsize 12
\spacing single
\use_hyperref false
\papersize a4paper
\use_geometry false
\use_amsmath 1
\use_esint 1
\use_mhchem 1
\use_mathdots 1
\cite_engine natbib_authoryear
\use_bibtopic false
\use_indices false
\paperorientation portrait
\suppress_date false
\use_refstyle 1
\index Index
\shortcut idx
\color #008000
\end_index
\secnumdepth 3
\tocdepth 3
\paragraph_separation indent
\paragraph_indentation default
\quotes_language english
\papercolumns 1
\papersides 2
\paperpagestyle default
\tracking_changes false
\output_changes false
\html_math_output 0
\html_css_as_file 0
\html_be_strict false
\end_header
\begin_body
\begin_layout Standard
\begin_inset CommandInset include
LatexCommand include
filename "portada.lyx"
\end_inset
\end_layout
\begin_layout Standard
\begin_inset ERT
status open
\begin_layout Plain Layout
\backslash
pagenumbering{roman}
\end_layout
\end_inset
\end_layout
\begin_layout Chapter*
Summary
\end_layout
\begin_layout Standard
This thesis has been developed as part of the Cajal Blue Brain research
project, whose long-term goal is simulating an anatomically correct human
brain inside a supercomputer.
One of the main areas of involvement of the UPM in this project is in the
development of new technologies for image analysis and visualization.
These new technologies are needed to extract useful information from 3D
images obtained with modern electronic microscopes from real cortical tissue.
This task presents a series of challenges, derived amongst other things
from the large amount of images that can be obtained from a small volume
of tissue as well as from the characteristics innate to this images.
Today, the analysis of the images is done by hand by trained scientists,
this represents a huge bottleneck for the project, as a well as a very
expensive and error prone process.
\end_layout
\begin_layout Standard
The objective of this master thesis is the development of an interactive
tool for segmenting a stack of brain tissue images.
This tool will use segmentation techniques based on curve evolution or
Snakes.
A major limitation of curve evolution algorithms is their robustness and
computational complexity, which prevents them from being used in an interactive
segmentation tool.
In this thesis we will implement two recent evolution algorithms proposed
by
\begin_inset CommandInset citation
LatexCommand citet
key "Marquez_2012"
\end_inset
, that are based on the use of morphological operators to overcome some
of the limitations of traditional Snakes evolution methods.
The result of the implementation is a plug-in for a neuronal analysis system,
called ESPina, which is being developed by various research groups in the
UPM.
In the thesis we also analyze the challenges that arise during the implementati
on of these algorithms as well as the improvements and opportunities that
this work opens.
\end_layout
\begin_layout Chapter*
\lang spanish
Resumen
\end_layout
\begin_layout Standard
\lang spanish
Esta tesis ha sido desarrollada como parte del proyecto de investigación
Cajal Blue Brain cuyo principal objetivo a largo plazo es el de realizar
una simulación anatómicamente correcta de un cerebro humano en un supercomputad
or.
Una de las principales áreas del proyecto en la cual está involucrada la
UPM es en el desarrollo de nuevas tecnologías de análisis y visualización
de imágenes.
Estas tecnologías son necesarias para extraer información útil de imágenes
en 3D obtenidas mediante técnicas modernas de microscopia electrónica aplicadas
a especímenes reales de tejido cortical.
Esta tarea presenta una serie de retos, derivados entre otras cosas por
el inmenso volumen de imágenes que se generan a partir de un volumen muy
pequeño de tejido así como por las características propias de las imágenes.
Hoy en día, este análisis se hace a mano por científicos entrenados, esto
representa además de un cuello de botella importante para el proyecto,
un proceso sumamente costoso y propenso a errores.
\end_layout
\begin_layout Standard
\lang spanish
El objetivo de esta tesis es el desarrollo de una herramienta interactiva
de segmentación de imágenes de tejido cerebral.
Esta herramienta utilizará tecnicas de segmentacion de imagenes basadas
en evolucion de curvas o Snakes.
Los algoritmos de evolución de curvas presentan limitaciones en cuanto
a robustez y complejidad computacional que limitan su utilidad en una herramien
ta interactiva.
En esta tesis implementaremos dos algoritmos propuestos recientemente por
\begin_inset CommandInset citation
LatexCommand citet
key "Marquez_2012"
\end_inset
, los cuales se basan en el uso de operadores morfológicos para superar
algunas de las limitaciones de los enfoques tradicionales de evolución
de Snakes.
El resultado de la implementación es un un plug-in para el sistema de visualiza
ción y análisis neuronal ESPina, el cual está siendo desarrollando por varios
grupos de investigación de la UPM.
En la tesis también analizamos los retos que se nos presentaron para la
implementación de los algoritmos así como las posibles mejoras y avenidas
de investigación que se abren a partir de nuestro trabajo.
\end_layout
\begin_layout Standard
\begin_inset CommandInset toc
LatexCommand tableofcontents
\end_inset
\end_layout
\begin_layout Standard
\begin_inset FloatList figure
\end_inset
\end_layout
\begin_layout Standard
\begin_inset ERT
status open
\begin_layout Plain Layout
\backslash
clearpage
\end_layout
\begin_layout Plain Layout
\backslash
pagenumbering{arabic}
\end_layout
\end_inset
\end_layout
\begin_layout Chapter
Introduction
\end_layout
\begin_layout Section
The problem
\end_layout
\begin_layout Standard
The Blue Brain Project, a collaboration between Switzerland's EPFL and IBM,
is a project with a bold objective.
Their goal is to make a virtual model of the human brain, down to the last
neuron, and host it on a supercomputer.
Needless to say, this kind of project has a very big set of challenges
to overcome.
The Technical University of Madrid (UPM) as well as some other Spanish
scientific institutions are strategic partners in this project and the
entity that represents these Spanish contributions is called Cajal Blue
Brain Project.
\end_layout
\begin_layout Standard
In order to obtain morphological information about the structure of the
brain directly from actual brain tissue modern electronic microscopy techniques
are being used to obtain highly detailed images from samples of cerebral
cortex.
These images are extremely useful because when properly analyzed by trained
scientist they can yield important quantitative as well as qualitative
data.
When done by hand, the processing of the images is a very time consuming
task that is labor intensive and therefore expensive.
Combine the cost (in every sense) of this process to the fact that the
amount of images that have to be processed is immense and you have a really
hard problem in your hands
\end_layout
\begin_layout Standard
One of the main techniques used for the extraction of meaningful data from
digital images, and that could be a possible solution for the problem at
hand, is image segmentation.
Image segmentation is the process of labeling or clustering the pixels
of an image according to their characteristics.
A good example of image segmentation would be to label all the pixels in
an aerial image depending if they are part of a road or not.
The result of the process would be that every pixel would be given a label
(road or not-road) depending on whether or not the pixel belongs to the
road.
When applied to a medical image, the proper segmentation of an image might
allow the user to obtain data, for example, about the amount of instances
of a given structure in a given area, the average shape or size of an interesti
ng feature, etc.
\end_layout
\begin_layout Standard
\begin_inset Float figure
wide false
sideways false
status open
\begin_layout Plain Layout
\begin_inset Graphics
filename Images/segment.png
display false
width 100col%
\end_inset
\end_layout
\begin_layout Plain Layout
\begin_inset Caption
\begin_layout Plain Layout
Example of a segmented image.
Image A is the original picture.
Image B is the result of the segmentation, with blue labeled pixels representin
g the road and red labeled pixels representing not-road.
\begin_inset Newline newline
\end_inset
[Figure created by the author, source of base image
\begin_inset Newline newline
\end_inset
\begin_inset Flex URL
status open
\begin_layout Plain Layout
http://www.flickr.com/photos/liao/2549120611
\end_layout
\end_inset
]
\end_layout
\end_inset
\end_layout
\end_inset
\end_layout
\begin_layout Standard
To summarize, the problem is the following: how can the scientists of the
Cajal Blue Brain segment images of brain tissue obtained with electronic
microscopes with the objective of differentiating several types of structures
within those cells (e.g.
synapses, vesicles, spines, dendrites) in the easiest, fastest and most
automated way?
\end_layout
\begin_layout Section
The solution
\end_layout
\begin_layout Standard
Given the previous description of the problem, we propose a solution based
on two elements.
First, an environment or framework on which the scientists can interact
with the images in a useful way and secondly an implementation of an algorithm
(or algorithms) that allows the scientist to segment the images in a time
efficient manner.
\end_layout
\begin_layout Standard
For the first part of the solution we are going to develop a plug-in for
a system called ESPina, created as part of the Cajal Blue Brain initiative.
This program allows a user to load a 3D stack of images and interact with
them in useful ways.
It also provides two key functionalities, first it was designed from the
ground up to support image segmentation and provides a system for categorizing
hierarchically the labeled structures and secondly and more importantly,
it provides an open architecture that can be enhanced with external plug-ins
which will allow us to build a tool to do segmentation interactively using
two selected algorithms.
\end_layout
\begin_layout Standard
The second component of the solution is the implementation of the two algorithms
mentioned before.
The segmentation problem has been around for quite some time and a lot
of approaches have been tried in solving it.
In the next chapter we provide an overview of those approaches and their
limitations.
Our solution is based on the use of active contours, also known as Snakes.
This approach is based on the idea of iteratively evolving a curve using
a gradient descent technique based on the minimization of some energy function.
The classical way of solving this problem is using Partial Differential
Equations (PDEs) but, as will be discussed later on, this process has some
limitations that can be solved by the use of certain morphological operators
that have been proposed recently.
The use of morphological operators, has the advantage of being simpler,
faster and numerically more stable than the their differential counterparts.
\end_layout
\begin_layout Standard
The ESPina tool, is built as an extension of ParaView, an open-source, multi-pla
tform data analysis and visualization application.
In order to be compatible with ESPina/ParaView, the implementation of the
segmentation algorithms has to be done using Kitware's Visualization Toolkit
(VTK) which is a free, open source software system for 3D computer graphics,
image processing and visualization created by the same company that created
ParaView, Kitware Inc..
\end_layout
\begin_layout Standard
One of the main objectives in the design of the segmentation tool is for
the user to be able to interactively control the segmentation process,
even the evolution of the active contour curve.
This approach allows us to overcome some of the limitations of the segmentation
algorithms by incorporating the knowledge of the experts in the process.
\end_layout
\begin_layout Chapter
Some context.
The Cajal Blue Brain Project.
\begin_inset CommandInset label
LatexCommand label
name "chap:Context"
\end_inset
\end_layout
\begin_layout Standard
In 2005, a neuroscientist named Henry Markram at Ecole Polytechnique Fédérale
de Lausanne (EPFL) joined efforts with IBM with a very ambitious objective,
they planned to build a perfect replica of the human brain inside a computer.
According to their website (
\begin_inset Flex URL
status open
\begin_layout Plain Layout
http://bluebrain.epfl.ch
\end_layout
\end_inset
) the objectives of this project were:
\end_layout
\begin_layout Enumerate
Create a Brain Simulation Facility with the ability to build models of the
healthy and diseased brain, at different scales, with different levels
of detail in different species.
\end_layout
\begin_layout Enumerate
Demonstrate the feasibility and value of this strategy by creating and validatin
g a biologically detailed model of the neocortical column in the somatosensory
cortex of young rats.
\end_layout
\begin_layout Enumerate
Use this model to discover basic principles governing the structure and
function of the brain.
\end_layout
\begin_layout Enumerate
Exploit these principles to create larger more detailed brain models, and
to develop strategies to model the complete human brain.
\end_layout
\begin_layout Standard
As the project advanced, a collaboration effort with several Spanish institution
s was forged (namely Madrid Technical University (UPM) and the Instituto
Cajal, part of the Consejo Superior de Investigaciones Cientificas (CSIC)).
This Spanish contribution to the project was designated Cajal Blue Brain,
in honor to the Nobel laureate Santiago Ramon y Cajal.
Ramon y Cajal received the Nobel Prize in Physiology or Medicine in 1906
for his discoveries in neuroscience and for setting up the foundations
of what would eventually become the "neuron theory".
\end_layout
\begin_layout Standard
According to the memory of the project, its Vision, Mission and Objectives
are:
\end_layout
\begin_layout Subsection*
Vision
\end_layout
\begin_layout Standard
Biologically accurate computer models of mammalian and human brains could
provide a new foundation for understanding functions and malfunctions of
the brain and for a new generation of information-based, customized medicine.
\end_layout
\begin_layout Subsection*
Mission
\end_layout
\begin_layout Standard
To gather all existing knowledge of the brain, accelerate the global research
effort of reverse engineering the structure and function of the components
of the brain, and to build a complete theoretical framework that can orchestrat
e the reconstruction of the brain of mammals and man from the genetic to
the whole brain levels, into computer models for simulation, visualization
and automatic knowledge archiving by 2015.
\end_layout
\begin_layout Subsection*
Objectives
\end_layout
\begin_layout Enumerate
Assemble all existing knowledge of the brain by providing a new model and
informatics platform as a service to "publish" their data and models.
\end_layout
\begin_layout Enumerate
Accelerate the process of reverse engineering the brain by stimulating and
facilitating neuroscientists to shift towards industrial-scale neuroscience.
\end_layout
\begin_layout Enumerate
Align experimentalists and theoreticians to further derive the structure
and function of the brain in a manner that allows the construction of computer
models for all components of the brain.
\end_layout
\begin_layout Enumerate
Construct a new generation data-driven, model-oriented, emergent and evolvable
and database-architecture that allows dynamic assembling of the components
of the brain into whole brain models that can be used to simulate the functions
and malfunctions of the brain and to explore new treatments.
\end_layout
\begin_layout Enumerate
Establish an international brain simulation capability to serve as a foundation
for future basic and clinical research and medical treatment.
\end_layout
\begin_layout Enumerate
Couple future brain research funding to the global industry that it produces
and supports.
\end_layout
\begin_layout Standard
This general definition of the project echoes the definition of the Internationa
l Blue Brain Projects that it belongs to.
\end_layout
\begin_layout Standard
To attain its objectives, the CBBP is organized according to two more tactical
axes:
\end_layout
\begin_layout Itemize
The anatomical and functional micro organization of the cortical column.
\end_layout
\begin_layout Itemize
The development of potentially transferable biomedical technology.
\end_layout
\begin_layout Standard
Of these technologies that have to be developed some of the must crucial
ones are image analysis and visualization technologies.
\end_layout
\begin_layout Standard
The use of modern electronic microscopy techniques (cross-beam type microscopes)
allows for the automatic obtention of highly detailed 3D images of brain
tissue samples, a very large number of these can be obtained very fast
using these new microscopes when compared to traditional ways of doing
image acquisition via electronic microscopy.
This images are obtained by serially imaging "cuts" of the brain tissue
(e.g.
every 25 nm) and then stacking this slices into 3D representations of the
sampled tissue.
These images have to be analyzed with the objective of obtaining morphological
data about the cerebral cortex.
\end_layout
\begin_layout Standard
\begin_inset Float figure
wide false
sideways false
status open
\begin_layout Plain Layout
\align center
\begin_inset Graphics
filename Images/brainTissue.png
display false
width 80text%
\end_inset
\end_layout
\begin_layout Plain Layout
\begin_inset Caption
\begin_layout Plain Layout
A slice of a brain tissue image stack obtained with a cross-beam microscope.
\begin_inset CommandInset label
LatexCommand label
name "fig:Brain-tissue"
\end_inset
\end_layout
\end_inset
\end_layout
\end_inset
\end_layout
\begin_layout Standard
The problem is that these images have to be analyzed by hand by trained
scientists that are capable of labeling the different structures within
the image and thus enable the quantification of, for example, the density
of synapses in a given volume.
It goes without saying that is a very expensive, error prone and time consuming
task that could be addressed by new advances in image segmentation and
analysis.
To give the reader an idea of the magnitude of the amount of data that
this task requires, in order to visualize a volume equivalent to 1 cubic
mm of brain tissue, we would have to process 35 petabytes of data (to put
this into perspective, as of this writing (2012) Google's worldwide servers
are reported to handle between 20 and 30 petabytes of data every day).
\end_layout
\begin_layout Standard
The images collected pose a number of challenges to be processed automatically
(noise, blurred edges, large number of objects), so new techniques will
have to be developed in order to achieve this goal.
\end_layout
\begin_layout Section
ESPina
\end_layout
\begin_layout Standard
The ESPina Interactive Neuron Analyzer, is a tool developed by the Cajal
Blue Brain Project to allow the scientists to visualize the images obtained
with the electronic microscopes.
\end_layout
\begin_layout Standard
The tool is built on top of ParaView, a parallel visualization framework.
It is designed as a client/server model that allows the use of processing
in a massive cluster (using MPI messaging technology), a single stand-alone
server or in a local machine.
It was designed around the workflow of the neuroscientists that are working
on CBBP and BBP, and so the visualization and interaction with the images
is done in way that it's natural for them.
\end_layout
\begin_layout Standard
ESPina provides some basic interaction and segmentation tools "out of the
box", but its greatest potential comes from the fact that it can be enhanced
by the use of plug-ins that can be built to include almost any new functionalit
y into the system (new segmentation tools, ways to extract information from
images, interaction techniques, etc.).
The tool that is being presented in this work is built as a plug-in for
ESPina, in order to harness all the functionality that it has already built
in as well as the familiarity that the neuroscientists already posses with
the system.
\end_layout
\begin_layout Chapter
Previous Work
\begin_inset CommandInset label
LatexCommand label
name "chap:Previous-Work"
\end_inset
\end_layout
\begin_layout Section
Image Segmentation
\end_layout
\begin_layout Standard
Image segmentation is the process of labeling the pixels of an image in
a way that allows us to group together pixels that share certain characteristic
s.
A common use of segmentation is to separate an object in an image from
its background, because of this, it is common to refer to the pixels that
are part of the structure that is being segmented as foreground pixels
and every other pixel as background pixels even on the cases when this
names do not make much sense.
The segmentation problem has been approached from several directions: Split
and merge techniques (
\begin_inset CommandInset citation
LatexCommand citet
key "Pavlidis_Liow_1990"
\end_inset
), normalized cuts (
\begin_inset CommandInset citation
LatexCommand citet
key "Shi_Malik_2000"
\end_inset
), graph cuts (
\begin_inset CommandInset citation
LatexCommand citet
key "Boykov_Jolly_2001"
\end_inset
), and active contours/Snakes (
\begin_inset CommandInset citation
LatexCommand citet
key "Kass_Witkin_Terzopoulos_1988"
\end_inset
,
\begin_inset CommandInset citation
LatexCommand citet
key "Malladi_Sethian_Vemuri_1995"
\end_inset
).
\end_layout
\begin_layout Standard
The algorithms we are going to use in our tool are based on the active contours
approach so we are going to discuss it in a little more detail.
Some examples of the utilization of Snakes for image segmentation can be
found in
\begin_inset CommandInset citation
LatexCommand citet
key "Nilsson_Heyden_2003,Chan_Vese_2001,Mortensen_Barrett_1995"
\end_inset
.
\end_layout
\begin_layout Standard
Active contour methods iteratively move towards a solution (segmentation)
under the combination of image and optional user guidance forces (
\begin_inset CommandInset citation
LatexCommand citet
after "p. 270"
key "Szeliski_2010"
\end_inset
).
The first active contour models relied on parametric curves, but they had
difficulty adapting the topology of the curve as it evolved.
As an alternative, level-set methods represent the curve as the zero-crossing
of a characteristic function (
\begin_inset CommandInset citation
LatexCommand citet
after "p. 281"
key "Szeliski_2010"
\end_inset
) and the evolution of the curve is done by modifying the underlying embedding
function.
We are going to analyze two models that use this representation:
\shape italic
Geodesic Active Contours
\shape default
(GAC) was proposed by
\begin_inset CommandInset citation
LatexCommand citet
key "Caselles_Kimmel_Sapiro_1997"
\end_inset
, the main idea in this model is that the evolution of the curve is dictated
by the minimization of an energy functional whose gradient dictates the
direction of movement for the curve, this energy functional is dependent
on a function (normally called the "stop criterion" and noted
\begin_inset Formula $g\left(I\right)\left(x\right)$
\end_inset
) thats takes low values in the areas of interest of the image.
Another model, originally proposed by
\begin_inset CommandInset citation
LatexCommand citet
key "Chan_Vese_2001"
\end_inset
, is called
\shape italic
Active Contours Without Edges
\shape default
(ACWE), it does not require a stop criterion function for minimizing its
energy functional, but instead relies in global information regarding the
contents (i.e.
the gray levels) of pixels inside and outside of the evolving curve.
In figure
\begin_inset CommandInset ref
LatexCommand ref
reference "fig:chan_vese_curve"
\end_inset
we can see a schematic representation of an evolving curve, notice how
the curve
\begin_inset Formula $C$
\end_inset
is represented by the zero level-set of the function
\begin_inset Formula $\phi$
\end_inset
.
\end_layout
\begin_layout Standard
\begin_inset Float figure
wide false
sideways false
status open
\begin_layout Plain Layout
\begin_inset Graphics
filename Images/ACC.png
display false
\end_inset
\end_layout
\begin_layout Plain Layout
\begin_inset Caption
\begin_layout Plain Layout
A curve being evolved using Active Contours Without Edges.
[Image source:
\begin_inset CommandInset citation
LatexCommand citet
key "Chan_Vese_2001"
\end_inset
]
\begin_inset CommandInset label
LatexCommand label
name "fig:chan_vese_curve"
\end_inset
\end_layout
\end_inset
\end_layout
\end_inset
\end_layout
\begin_layout Standard
Both of these approaches require the calculation of partial differential
equations which can be computationally costly, add complexity to the implementa
tion, can introduce numerical instabilities and require frequent reinitializatio
ns of the level-set function as it degrades over time.
\end_layout
\begin_layout Standard
As a proposed solution to this shortcomings of the traditional approach
to snakes evolution
\begin_inset CommandInset citation
LatexCommand citet
key "Alvarez_2010"
\end_inset
and
\begin_inset CommandInset citation
LatexCommand citet
key "Marquez_2012"
\end_inset
propose a method that replaces the terms of the PDEs in traditional algorithms
with infinitesimally equivalent morphological operators.
Using these new operators the representation of the curve is simpler and
the operations required to obtain it become less costly.
\end_layout
\begin_layout Subsection
Morphological Geodesic Active Contours (MGAC)
\begin_inset CommandInset label
LatexCommand label
name "sub:Morphological-Geodesic-Active"
\end_inset
\end_layout
\begin_layout Standard
As seen in
\begin_inset CommandInset citation
LatexCommand citet
key "Alvarez_2010"
\end_inset
and
\begin_inset CommandInset citation
LatexCommand citet
key "Marquez_2012"
\end_inset
, it is possible to approximate the behavior of the PDE that controls curve
evolution in the traditional GAC as the composition of three morphological
operators.
In general the PDE can be divided in three components, one represents the
attraction applied to the curve by the areas of interest in the image (e.g.
edges), one represents the balloon force that ensures that the curve does
not get stuck in parts of the image with little information and the third
component is the mean curvature operator which can be seen as a smoothing
force on the curve.
\end_layout
\begin_layout Standard
The PDE that we will approximate using the morphological operators is:
\end_layout
\begin_layout Standard
\begin_inset Formula
\begin{equation}
\frac{\partial u}{\partial t}=g\left(I\right)|\nabla u|\text{div}\left(\frac{\nabla u}{|\nabla u|}\right)+g\left(I\right)|\nabla u|\nu+\nabla g\left(I\right)\nabla u\label{eq:pde_gac}
\end{equation}
\end_inset
\end_layout
\begin_layout Standard
Where
\begin_inset Formula $u$
\end_inset
represents the embedding function,
\begin_inset Formula $g\left(I\right)$
\end_inset
represents a function that has low values in the areas of interest of the
image and
\begin_inset Formula $\nu$
\end_inset
is a parameter that controls the effect of the balloon force
\end_layout