The First National Conference on Electronics and New Technologies (NCENT’2015)
May 19-20, M'Sila, Algeria
1
Morphological Characterization of SEM Thin Silver Films Using Thresholding
Boudour Samah, Dehchar Charif
Welding and NDT Research Centre (CSC) Cheraga, Algiers, Algeria Thin Films and Applications Unit (UDCMA) – Setif
[email protected], [email protected]
Abstract—Thin Silver films given the great well-developed in the materials science and characterization technologies are in high requirement, given the wide-spread use of coatings in many engineering and science fields. In this paper, we report a simple image thresholding method to segment and to distinguish the pores from solid of thin Silver films prepared by electroless- plating deposition method at various plating times. This approach aims to analyze the detected pores on the surface of the studied samples by measuring their porosity, pores size and their distribution.
Keywords-component; binary image; image processing; image thresholding; porosity; segmentation; SEM image; thin silver film.
I. INTRODUCTION
Scanning Electron Microscopy images (SEM) of thin Silver films can be used to detect details of interest and to find morphological characteristics of these detected details [1].
Segmentation is a basal task in image processing and computer vision that can be used to segment and to distinguish the interesting details like pores, grains and others from SEM images [2]. Image thresholding is one of the basic techniques for segmentation. Take advantage of grays level image, thresholding can be used to create a binary image in black and white [2,3]. A threshold is set to tell the interesting detail biased from the rest of the image. Threshold moving average is the most classical method that can be used to calculate automatically this threshold. The features and characteristics of thresholded region such as its size, orientation and other parameters can be described by description phase.
SEM images in Fig. 1 prepared for treatment represent coatings of Silver metallic deposited by electrochemical process (electroless plating) on epoxy glass substrate surface.
In this process of preparation, two electroless plating immersion times have been considered, namely 10 and 20 minutes [4].
The second section presents filtering and segmentation tasks that can be used to improve interested details and extract them.
The third section presents the description method and its interesting basic mathematic estimation for thin Silver film characterization.
From the software side (Matlab), it will crop SEM images prepared to invalidate the description band and make them square.
Figure 1. Thin Silver film images at two immersion durations: (a) 10 minutes and (b) 20 minutes (area framed by red dsigne the cropped area).
II. IMAGE PROCESSING A. Image
Computer specialists sample the image to small elements named pixels. Fig. 2 shows that each in his own coordinates pixel has a gray level intensity restricted between [0 255] and mediates eight neighbors [2].
Figure 2. Image sampled to 5*5 pixels with its gray levels.
(a)
(b)
The First National Conference on Electronics and New Technologies (NCENT’2015)
May 19-20, M'Sila, Algeria
2 B. Image Filtering
The filtering is an operation neighborhood that emphasizes certain details or removes other details. We apply a median filter to adjust unevenly distributed gray values like noise and sharp discontinuity of intensity. The intensity altered is smoothed by Gaussian and Derich filters successively. The filtering results of two previous thin Silver films are shown in the following figures (Fig. 3) [2].
Figure 3. Thin Silver images filtered by median and Gaussian- Deriche filters ((a) 10 minutes and (b) 20 minutes).
C. Image Thresholding (Segmentation Task)
Moving average is a local thresholding segmentation which is based on computing a moving average along scan lines of an image. Suppose the 5*5 neighborhood presented in Fig. 2. The estimated threshold (mean) is different for each pixel and it was calculated by averaging its 4 preceding neighbors and itself [2].
The calculated threshold tells the pore pixels apart from the rest of the image according to the following expression [2]:
( ) { ( ) ( )
( ) ( )
When
f (x,y) is input image
g(x,y) is segmented output image
m(x,y) is the local average And K is a constant [0 1].
After applying the thresholding operation, all pore pixels in the image are converted to black (expressed mathematically by 0) and all solid pixels are converted to white (expressed mathematically by 1). Fig. 4 displays the segmented result of two previous filtered images shown in Fig. 3.
Figure 4. Segmented images of thin Silver films ((a) 10 minutes and (b) 20 minutes).
III. POROSIY DESCRIPTION
From Fig. 4, we notice that each sample contains i pores.
Each pore (i) can be described morphologically by its size (Si) and radius (ri).
Figure 5. Three pores and its segmented image (zomed).
The previous expressed parameters can be calculated from aggregate black pixels of each segmented pore according to the following equations:
∑ ∑ ( )
√ ⁄
(a)
(b)
(a)
(b)
The First National Conference on Electronics and New Technologies (NCENT’2015)
May 19-20, M'Sila, Algeria
3
∑
IV. RESULTS AND DISCUSSIONS
The results cited in the following graph and table is simulated in Matlab software. They represent the characterizations of thin Silver films AG1 at time immersion of 10 minutes and AG2 at time immersion of 20 minutes.
Figure 6. Pores radiuses (µm).
The graph displays two curved assign the number of pores to denote the radius. For example AG1 has 95 pores with length radius 0.2µm and AG2 has about 68 pores with the same radius subsequently AG1 has narrowed pores more than AG2.
AG1 has pores with long radius 1.7µm, 1.8µm and 2.2µm and the long radius in AG1 measure 1.6µm.
From Table. I, AG2 has a maximum radius which measure 2.20µm greater than AG1 which measure 1.60µm.
The mean radius in AG2 measure 0.55µm is greater than the mean radius of AG1 which measure 0.42µm. Also, the size of all pores of AG2 measure 384.72µm2 is greater than the size of all pores of AG1 which measure 270.08 compared to the size of whole image which measure 6400µm2. Contrary to described results cited previously, the number of pores of AG1 is more than the number of pores of AG2.
TABLE I.
measures Table pore radius
AG1 AG2 unit
Radiuses_max 1.60 2.20 µm
Radiuses_means 0.42 0.55 µm
Nbr_pores 371 299 /
Size_all_pores
6752 9618 pixel
270.08 384.72 µm2
Porosity or
Percentage_pores 3.22 6.01 %
Finally the porosity of AG2 which measure 6.01% of the whole image is greater than the porosity of AG1 which measure 3.22%.
V. CONCLUSION
By taking advantage of the digital image processing parts like filtering, segmentation and description, we apply a programmable methodology that generates specific parameters that will allow the materialists to characterize their SEM images of thin films.
REFERENCES
[1] S. Buchholz, H. Fuchs and J. P. Rabe , “Surface structure of thin metallic films on mica as seen by scanning tunneling microscopy, scanning electron microscopy, and lowenergy electron diffraction”, J. Vac. Sci. Technol. B, Vol. 9(2), 1991
[2] R.C. Gonzalez and R.E. Woods, Digital Image Processing , Pearson- Prentice-Hall, Third Edition, Location, 2008.
[3] M. Sezgin, and B. Sankür, “Survey Over Image Thresholding Techniques And Quantitative Performance Evaluation Journal Of Electronic Imaging”, Journal of Electronic Imaging, Vol. 13(1), pp.
146-165, January 2004.
[4] C. Dehchar, “Elaboration D’Electrodes Modifiées par Metallisation de Matériaux Non-Conducteurs”, Mémoire de Magister, Ecole Militaire Polytechnique, 2012.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 2 2.1 2.2 0
5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 90 95 100
pore radiuses
radius (per µm)
numbre of pores
AG1 (immersion time = 10 min) AG2 (immersion time = 20 min)
