Professional Documents
Culture Documents
Characterization of Bio-Dynamic Speckles Through Classical and Fuzzy Mathematical Morphology Tools
Characterization of Bio-Dynamic Speckles Through Classical and Fuzzy Mathematical Morphology Tools
3
Signal Processing
5
11
Characterization of bio-dynamic speckles through classical and fuzzy
13 mathematical morphology tools
15 Q1 Eduardo Blotta a,n, Agustina Bouchet a,b, Marcel Brun a, Virginia Ballarin a
a
17 Grupo de Procesamiento Digital de Imágenes, Facultad de Ingenierı́a, U.N.M.D.P. Mar del Plata, Argentina
b
Q3 Consejo Nacional de Investigaciones Cientı́ficas y Técnicas (CONICET), Argentina
19
21 a r t i c l e i n f o abstract
23 Article history: In this paper we characterize dynamic speckle signals, obtaining selective information
Received 2 May 2012 through the differentiation of morphological patterns of the temporal history of each
Received in revised form pixel, using the morphological granulometric function. This method is applied to the
25 26 October 2012
analysis of images of apples and corn seeds. Studies on the first ones were focused on
Accepted 4 January 2013
the activity on their surface, related to healthy and damaged areas, while for seeds on the
27 viability of the embryo and endosperm. Subsequently, the analysis was repeated using
Keywords: fuzzy mathematical morphology techniques, comparing the results obtained by both
29 Dynamic speckle methods.
Mathematical morphology
& 2013 Published by Elsevier B.V.
Fuzzy mathematical morphology
31 Morphological granulometric function
33 63
35 65
1. Introduction the morphological spectrum obtained by morphological
37 67
granulometry [4,3,5].
When a surface that has physical or biological activity This technique is based on the theory of mathematical
39 69
is illuminated by coherent light, such as a laser, the morphology (MM) [14,17], which is a powerful set of tools
scattered light displays a granular structure, i.e. present- for digital image processing. These techniques allow
41 71
ing randomly distributed light and dark spots, which enhancement of diffuse areas, object segmentation, edge
change over time, giving a visual effect of ‘‘water boiling’’. detection and structures analysis via the use of operators or
43 73
This effect is known as ‘‘dynamic’’ speckle or ‘‘bio- filters. This discipline is based on sets algebra and can be also
speckle’’. This phenomenon has been studied in various be used as a tool for signals analysis. The techniques, initially
45 75
types of biological samples, such as fruits [15] and seeds developed for binary images, were also extended to the field
[7]. Various methods and techniques have been developed of gray level images. Another approach to extend binary
47 77
for measuring the activity of a speckle pattern [8,1], but, operators to gray level images, the fuzzy mathematical
in most cases, the results consist of a single image, with morphology (FMM), is based on fuzzy sets theory. Often,
49 79
no analysis of the temporal information. Subsequently, the fuzzy operators result more robust than classical mor-
selective filtering techniques were used for studying the phological operators [6].
51 81
bio-speckle signals through spectral decomposition When illuminated by a laser light, fruits show an
[16,11]. Going in the same direction, we introduced speckle activity that can be related to the degree of
53 83
recently a new method for bio-speckle analysis based on maturity, turgor pressure, damage, age and mechanical
properties. For this reason, it is interesting to develop
55 85
numerical methods which allow to extract useful infor-
mation of the speckle image sequences.
57 87
n
Corresponding author. Tel.: þ54 223 474 5994. On the other hand, the study of the feasibility of germina-
Q5 Q4
E-mail address: eblotta@fi.mdp.edu.ar (E. Blotta). tion is an important topic in production and marketing seeds.
59 89
0165-1684/$ - see front matter & 2013 Published by Elsevier B.V.
61 http://dx.doi.org/10.1016/j.sigpro.2013.01.001
Please cite this article as: E. Blotta, et al., Characterization of bio-dynamic speckles through classical and fuzzy
mathematical morphology tools, Signal Processing (2013), http://dx.doi.org/10.1016/j.sigpro.2013.01.001i
2 E. Blotta et al. / Signal Processing ] (]]]]) ]]]–]]]
1 They have been developed many tests to determine strength Typical speckle signal 63
and viability of seed germination. Other studies about the 200
3 reliability of the tests and equivalence between them also 65
exist [12]. In this context, the evaluation of laser interfero- 150
5 metry techniques as tools for seeds analysis is worthy to take 67
Gray level
into account.
7 In this paper, we apply MM and FMM filters to biological 100 69
samples of red delicious apples, as a tool for the diagnostic of
9 early damage of the surface of the fruit, and to corn seeds, to 50
71
study its performance in the germination process.
11 73
0
50 100 150 200
13 75
t
2. Methods
15 Fig. 2. Typical temporal evolution of a bio-speckle. 77
2.1. Dynamic speckle
17 79
When a surface, that has a certain physical or biological
19 activity, is illuminated by a high coherence light beam, the 81
PC
scattered light by the surface has a granular structure +
21 composed of small bright and dark areas, randomly dis- Frame Grabber 83
tributed, which change over time, producing an visual
23 effect as a boiling liquid. 85
Fig. 1 shows a typical image. This effect, known as
25 ‘‘dynamic speckle’’, is a result of coherent light scattering 87
Video
by objects that exhibit some level of activity. These type Camera
27 of images present variations in the local intensity corre- (CCD) 89
sponding to the level of biological activity in the area
29 under observation. Fig. 2 shows the temporal evolution of 91
the intensity of a sample pixel, or gray levels, in a typical
31 sequence of dynamic speckle images. Due to the stochas- Laser 93
tic nature of the signal, it would be impossible with the
33 naked eye recognize the correspondence of this with any 95
particular area of a biological sample.
35 The dynamics of the speckle effect is usually quite Sample 97
complex due to multiple physical mechanisms involved
37 [18], but the activity evaluation can help to recognize the 99
Fig. 3. Experimental speckle bank.
complex processes occur in a biological sample.
39 101
2.2. Experiments
41 103
The test bank used to obtain the images is shown in
43 Fig. 3. We used a low-power He–Ne laser (5 mW, 105
l ¼ 633 nm) to illuminate the samples, using a divergent
45 beam expanded to encompass a wide region. Subjective 107
speckle images were formed by a objective (usually
47 f¼ 50 mm, f/#¼16). Thus, measurement of average 109
speckle grains covered several pixels. For the experiment
49 with apples, an inert reference object was added in the 111
images. Successive images were stored on a PC by a CCD
51 camera connected to a acquisition board. Low lighting 113
levels were used so that the effect of irradiation on the
53 sample was negligible. The laser illumination was 115
adjusted to keep constant the average intensity in the
55 image throughout the test. 117
To study variations of the speckle phenomenon in the
57 surface of the fruit, a controlled hit was applied to the 119
healthy apples. That hit was caused by a falling steel ball
59 (dm¼21.9 mm, weight¼33.6 g) from a height of 20 cm of 121
fruit surface. The damage in the sample could not be
61 Fig. 1. Typical speckle pattern. appreciated by simple visual inspection. Images were 123
Please cite this article as: E. Blotta, et al., Characterization of bio-dynamic speckles through classical and fuzzy
mathematical morphology tools, Signal Processing (2013), http://dx.doi.org/10.1016/j.sigpro.2013.01.001i
E. Blotta et al. / Signal Processing ] (]]]]) ]]]–]]] 3
1 acquired in 300 300 pixels, digitized into 256 gray Speckle signal: before and after aperture filtering 63
levels, each 0.5 s approximately. Sequences took 500 200
3 pictures before, and immediately after the coup. 65
Described image sequences were stored in 3D arrays
150
5 (300 300 256) where the third variable is the time. 67
Gray level
Thus, the temporal evolution of each pixel can be ana-
7 lyzed to detect changes in biological activity due to 69
100
changes in the surface of the fruit.
9 In the case of seeds, experiments cited by the original 71
paper [7] were held to explore how moisture affects
50
11 speckle activity. This work, however, was focused on the 73
qualitative aspects of the samples and the different 100 120 140 160 180 200
13 morphologies in order to segment the interest areas of t 75
the seed, for analysis. We used specimens of corn seeds, Fig. 4. Morphological opening applied on a bio-speckle signal.
15 previously wetted, cutted in half and we took sequences 77
of 100 images of 256 480 pixels, using the bank
17 previously described, with the same configuration. GSD of a bio−speckle signal 79
0.8
Ø(n)
morphology (MM) are the dilation and the erosion. Many 0.4
23 other operators of the MM are built combining these two 85
operators. 0.2
25 We will describe in this section the morphological 87
operators used for the analysis of the unidimensional bio-
27 0 89
speckle signals, defined by f : D Z-f0,1,2, . . . ,255g, 0 20 40 60 80 100
where Z is the set of integer numbers. n
29 The dilation of the signal f by the signal b, called also 91
structuring element (SE) is defined by [17] Fig. 5. Example of the granulometric size distribution for a bio-speckle
31 signal. 93
dðf ,bÞðxÞ ¼ supff ðxyÞ þ bðyÞg ð1Þ
y2Db
33 is probed by increasing shapes (the SE), which can be 95
where x 2 Df , and Df and Db are the support of f and b, compared to a selective filtering of the shape of the signal,
35 respectively. since the signal is smoothed with more strength in each 97
The same way we define the erosion of the signal f by iteration. The operation ends for a value n ¼N, which
37 the signal b, by [17] depends on the shape of the signal, when there is no 99
eðf ,bÞðxÞ ¼ inf ff ðxyÞbðyÞg ð2Þ variation between two consecutive openings.
y2Db
39 Fig. 5 shows an example of GSD calculated for the 101
Based on these two operators of dilation and erosion, signal of Fig. 4.
41 the opening is defined as the combination of a erosion 103
followed by a dilation:
43 2.4. Fuzzy mathematical morphology 105
gðf ,bÞ ¼ dðeðf ,bÞ,bÞ ð3Þ
45 Both, the operator and the SE chosen, will determine The fuzzy mathematical morphology (FMM) provides a 107
the characteristics of the resulting processed signal. way to extend mathematical morphology’s binary operators
47 Fig. 4 shows the effect on the signal obtained by the to gray level images, extending the set operations to fuzzy set 109
application of a morphological opening to a bio-speckle ones, based on the theory of fuzzy sets. Fuzzy sets are defined
49 signal, with a linear SE of length 2. by their ‘‘membership’’ function, with values in the interval 111
If we consider Oðf 0 Þ the area under the original signal, [0,1] [10].
51 and Oðf n Þ the area under the signal fn, obtained by For the unidimensional case, gray level signals with 113
applying an opening to f0 with a lineal SE of length n, range between 0 and 255 can be ‘‘fuzzified’’ to fit in the
53 we can define the granulometric size distribution fðnÞ, or [0,1] range, via the function g : f0,1,2 . . . ,255g-½0,1 115
GSD [9], by x
55 gðxÞ ¼ ð5Þ 117
Oðf n Þ 255
fðnÞ ¼ 1 , n ¼ 0, . . . ,N ð4Þ
Oðf 0 Þ
57 It is important to note that this does not mean that the 119
fðnÞ is a non-decreasing function, similar to statistical signal represents a fuzzy membership function for some
59 cumulative distribution, but without reaching the value object, this is just a way to model the gray level signal 121
one, which is fixed later by a normalization factor. It to be able to apply morphological operators defined via
61 represents the variation of the area under the signal as it fuzzy set operations. Signals can be defuzzified by applying 123
Please cite this article as: E. Blotta, et al., Characterization of bio-dynamic speckles through classical and fuzzy
mathematical morphology tools, Signal Processing (2013), http://dx.doi.org/10.1016/j.sigpro.2013.01.001i
4 E. Blotta et al. / Signal Processing ] (]]]]) ]]]–]]]
Please cite this article as: E. Blotta, et al., Characterization of bio-dynamic speckles through classical and fuzzy
mathematical morphology tools, Signal Processing (2013), http://dx.doi.org/10.1016/j.sigpro.2013.01.001i
E. Blotta et al. / Signal Processing ] (]]]]) ]]]–]]] 5
1 63
3 65
5 67
7 69
9 71
11 73
13 75
15 77
17 79
19 81
21 83
23 85
25 87
27 89
29 91
31 93
33 95
35 97
F F F
Fig. 6. Apple images after morphological filtering (SE lengths: 2, 10 and 30): (a)–(c) fð2Þ, fð10Þ and fð30Þ, respectively. (d)–(f) f ð2Þ, f ð10Þ and f ð30Þ,
37 F F F
using Dubois & Prade norms. (g)–(i) f ð2Þ, f ð10Þ and f ð30Þ, using standard norms.
99
39 101
Table 3
41 Mean of square quadratic errors between gold standard apple image and 103
filtered speckle image. Lower values mean better segmentation
performance.
43 105
Granulometry SE ¼ 2 SE ¼10 SE ¼ 30
45 107
Classic 0.5356 0.3978 0.7796
Dubois & Prade 0.7691 2.0664 2.2075
47 Standard 0.7691 2.9491 2.9472
109
49 111
hypocotil, and it is the first structure to raises from the
51 soil during germination. 113
Fig. 8, shows the results of the GSD for on the speckle
53 signal for corn, using three different SEs, (a)–(c), and the 115
results of the FGSD for the same signal and SEs, (d)–(f),
55 using the Dubois & Prade norms, and finally the results of 117
Fig. 7. Constitutive parts of a corn seed.
the FGSD using the standard norm, (g)–(i).
57 Fig. 8 shows the results for corn. In images (a)–(c), 119
obtained by using MM, we can see a major contrast Table 4 shows the MSE computed for corn seed,
59 between the different regions, while images (d)–(i), contrasted to a hand-drawn image, with optimal segmen- 121
obtained by FMM (Dubouis & Prade and standard norms), ted corn regions of embryo and endosperm. Lower values
61 show uniform gray levels in the embryo region. mean better segmentation performance. 123
Please cite this article as: E. Blotta, et al., Characterization of bio-dynamic speckles through classical and fuzzy
mathematical morphology tools, Signal Processing (2013), http://dx.doi.org/10.1016/j.sigpro.2013.01.001i
6 E. Blotta et al. / Signal Processing ] (]]]]) ]]]–]]]
1 63
3 65
5 67
7 69
9 71
11 73
13 75
15 77
17 79
19 81
21 83
23 85
25 87
27 89
29 91
31 93
33 95
35 97
37 99
39 101
41 103
43 105
45 107
47 109
49 111
51 113
53 115
55 117
57 119
59 121
F F
Fig. 8. Corn seed images after morphological filtering (SE lengths: 2, 10 and 30): (a)–(c) fð2Þ, fð10Þ and fð30Þ, respectively. (d)–(f) f ð2Þ, f ð10Þ and
61 fF ð30Þ, using Dubois & Prade norms. (g)–(i) fF ð2Þ, fF ð10Þ and fF ð30Þ, using standard norms. 123
E. Blotta et al. / Signal Processing ] (]]]]) ]]]–]]] 7
49
Please cite this article as: E. Blotta, et al., Characterization of bio-dynamic speckles through classical and fuzzy
mathematical morphology tools, Signal Processing (2013), http://dx.doi.org/10.1016/j.sigpro.2013.01.001i