Sie sind auf Seite 1von 4

2007 International Conference on Computational Intelligence and Security Workshops

Research and Realization of the Computer-assisted Colorimetry for Baked


Porcelain Tooth

Jihong Liu Fangfan Han


College of Information Science and Sino-Dutch Biomedical and Information
Engineering, Northeastern University, Engineering School, Northeastern University,
Shenyang,110004 Liaoning, China Shenyang,110004 Liaoning, China
Brain Science Research Center, Korea neusiluhuayu163@sina.com
Advanced Institute of Science and Technology,
Daejeon,305-701, Republic of Korea
liujihong@ise.neu.edu.cn

Abstract the differences of the natural tooth’s color isn’t obvious,


the operators will hardly decide which color is best.
The defect of dentition and its abnormal and tooth The Computer-aided Colorimetric System for Baked
row imperfection are the common diseases. Along with Porcelain Tooth (CaCSBPT) makes use of the
the development of science and technology, the dental computer and relative software[1]. It determines the
restoration has already not only been limited to match color value of the compared tooth, and chooses the best
in the shape, but also been requested the color matched matched model in the baked porcelain tooth model sets.
with the natural tooth perfectly, make the dental case The colorimetric algorithm for the baked porcelain
more lifelike on the sense of vision. The computer color tooth is the core in the whole system.
matching (CCM) can satisfy this requirement. It can
also avoid the effect of the surroundings to artificial 2. The frame of CaCSBPT
colorimetric and improve the efficiency. After the
analysis of tooth color character, three kinds of CaCSBPT is used to choose the proper color for the
algorithms are adopted to carry on the colorimetry for restored tooth assisted by computer[2].
baked porcelain tooth, which are the center colorimetry, In this work, using Vita shade guide (16 shade tabs)
the nine sub-areas substraction colorimetry and the as a template, the images of 16 tooth models are
improved algorithm. Through analysis the accuracy collected by digital cameras and built as a teeth model
and time consuming in colorimetry, the advantages and set. After cropping the images and extracting the color
disadvantages of the three algorithms are compared. features in the set, the color future information is used
The Computer-aided Colorimetric System for Baked to build the model database. After the system gets tooth
Porcelain Tooth(CaCSBPT) is also researched. This sample image of the compared tooth by image capture
design is accomplished in Matlab platform. According part, it compares its color feature with that in the
to the reflection from clinicians, this design has better database and then chooses the best matching type.
clinical value. Figure1 shows the frame of CaCSBPT.

Obtain the tooth Extract the color characters


sample image of the target image
1. Introduction
With the development of the dental restoration Colorimetric Output the dental
technology, porcelain crown and bridge has become the algorithm colorimetric result
major restoration measure. The color matching of the
baked porcelain tooth is the most important step during Tooth model database
the process of porcelain tooth creation. Due to the
colorimetric method, surroundings and operators are Figure1.The functional sketch of colorimetric system
generally different, using the naked eye often cause the
colorimetric result very discontented, especially when

0-7695-3073-7/07 $25.00 © 2007 IEEE 128


DOI 10.1109/CIS.Workshops.2007.8
3. The design of colorimetric algorithm image separately as the objects for research[5]. Firstly, it
divides the cropped tooth image into nine sub-areas,
3.1 General description chooses the center sub-area to be analyzed, and
compares the three characteristic values with that in
The Vita16 shade guide, made by Japanese database, finally finds the best matched model. Figure
company Shofu, is used as the baked porcelain dental 4 shows the nine divided sub-areas of the cropped tooth
model in this design. image. V is the center sub-area[6].
According to the understanding and analysis of the
lamp-house and the comparison of images taken at
various irradiation source, the dental models are
irradiated under natural light (from 10:00 to 14:00)
with 45 degrees on the uniform blue background, and
images are taken by digital camera parallelly. There is
almost no shadow in the images and are convenient for Figure 4. The nine divided areas of the cropped tooth image
experiments. A group of images of shade guide are According to the pixel points of the center area, the
shown in Figure 2. averages of the tooth sample and the 16 tooth models
are obtained as Hs, Hm1~Hm16, Ss, Sm1~Sm16, Is, Im1~Im16.
In accordance with formula (1), the subtractions of
three color factors of teeth samples and each type tooth
model are obtained as ∆1~∆16.
∆ i = H s − H mi + Ss − S mi + Is − I mi ( i = 1,2," ,16 ) (1)
Finally, the corresponding model getting the
smallest ∆i is the most matching model with the tooth
sample.

3.3 The nine sub-areas subtract colorimetry

In this algorithm, each cropped tooth image is


Figure2.A group of photos of shade guide divided into nine sub-areas, of which H, S, I values are
In addition, after investigating the principle of compared separately. After considering the subtraction
CaCSBPT[3] and analyzing the dental color of color characters in nine areas, the corresponding
characteristics, it is found that the color characters in model number with the smallest value is found[7].
the largest internal rectangular in a tooth are studied to Based on the minimum distance algorithm,
stand for the tooth’s color features in the paper, as assuming that the hue averages of the nine areas of the
shown in Figure3. tooth sample image are H1~H9, that in a model are
Hm1 ~Hm9, the hue subtraction of nine areas of the
tooth sample and the model are gained according to the
formula (2). Then the 16 hue subtractions ∆Hi
(i=1,2,…,16) are gained by subtracting the hue
averages in nine areas of 16 models from that of the
tooth sample separately.
∆H = H1 − H m1 + H 2 − H m2 +" + H 9 − H m9 (2)
Figure3.The schematic study ∆S = S1 − S m1 + S 2 − S m2 +" + S9 − S m9 (3)
Consider the good performance of HSI color model,
it is used in this design. Three color characters values, ∆I = I1 − I m1 + I 2 − I m2 + " + I9 − I m9 (4)
Hue, Saturation and Intensity[4], for each tooth model Similarly, according to formula (3) and (4), the
are extracted to be used as the feature values, after subtractions of saturation (∆Si) and intensity (∆Ii) are
cropping teeth sample images with the mouse. gained, when i = 1, 2, ..., 16.
Finally the subtraction of three color factors
3.2 The center colorimetry between the teeth sample and each type tooth model, ∆i,
is computed by summation(i = 1, 2, ..., 16), when
The design principle of the center colorimetry is to compared the 16 values to get minimum value, the
extract the H, S and I image from the tooth model corresponding model is gained which is the most

129
matching model with the tooth sample. The third function is used to build the database of
the teeth models. It inputs the pictures of shade tabs of
3.4 The comparison of the two algorithms Vita16 one by one and calculates the nine characteristic
groups. There are single H, S and I value in every
The same points of the center area colorimetry and group. At last, they are saved in a data file called model
the nine sub-areas subtract colorimetry: database as 16 matrixes with 9×3 elements each. The
(a) Cropping teeth with the mouse to get the largest data file will be called by the main function when
rectangular area of the regional image. needed.
(b) RGB images are all converted into HSI images, and The forth function is the main function which
H, S, I images are extracted for research. calculates the characteristic averages in the nine
(c) The research objects are all divided into nine areas. sub-areas of the tooth sample to be compared. Then it
(d) H, S, I subtractions are all considered for compares the characteristic values of the tooth sample
comparison calculation. to the values of every tooth model in the model
The differences between the center area colorimetry database. Finally, according to the formulas (5), (6), (7)
and the nine sub-areas subtract colorimetry: and (8), the minimum ∆j is calculated, and then the
(a) The first algorithm considers only the H, S, I corresponding model is gained which is the best
information in the center area, but another is under the matching model with the tooth sample. The main
comprehensive consideration of H, S, I subtractions in function flow chart is shown in Figure 5.
nine areas.
(b) The computation of the second algorithm is more Begin
complex than that of the first.
Input the image of the tooth sample
3.5 The improved algorithm

Assuming that the color character values of the teeth Crop the image and calculate the color
samples are H1~H9、S1~S9、I1~I9, that in j sample in character values of the nine areas

the model set are Hj1 ~Hj9 、Sj1 ~Sj9 、Ij1 ~Ij9, the
subtractions of color character values between them are Read the color character
gained according to the formula(5) 、(6)、 (7)and values of one tooth model
(8), then ∆j is computed.
(H − H
1 j1
)2 + (H − H
2 j2
) 2 + " + ( H 9 − H j9 ) 2 (5) Calculate the subtraction according
∆hj =
9 to formula (5), (6), (7) and (8)
(S
1
− S
j1
)2 + (S
2
− S
j2
)2 +" + (S9 − S j9 )2 (6)
∆ sj =
9
N
(I − I )2 + (I − I )2 +" + (I9 − I )2 (7) Is it smaller than
1 j1 2 j2 j9
∆ ij = the last one?
9
∆ j = ∆ hj + ∆ sj + ∆ ij (8) Y
The smallest ∆j gained corresponding model is most
matching model with the tooth sample. Assign the serial number of
the current model to L
There are four functions in the programming of this
algorithm.
The first function is used to calculate the color N
characteristic averages(H, S and I).It inputs H, S or I Have the 16
image and divides them into nine sub-areas, then models finished?
calculates the color characteristic average of each
Y
sub-area through the number of pixels.
The second function is used to crop the picture and L value is the best matching model
calculate the averages. It inputs a RGB picture of the number in the model database
tooth, then crops it to achieve its internal largest
rectangular area, then converts it to HSI image. H, S
and I images are extracted from the HSI image, and End
then the characteristic averages in the nine sub-areas of
the input sample are achieved by calling the first Figure 5. The main function flow chart
function.

130
3.6 The analysis of experimental results 4 Conclusion
This related algorithms and interface are After investigating the principle of CaCSBPT and
accomplished in Matlab platform. In this paper, three analyzing the dental color characteristics, three kinds of
experiments are carried on in three algorithms, algorithms are designed under the Matlab platform to
experimental results are shown in Table 1.Where, I carry on the colorimetry for baked porcelain tooth,
stands for the center area colorimetry; II stands for the which are the center area colorimetry, the nine
nine sub-areas subtract colorimetry; and III stands for sub-areas subtract colorimetry and the improved
the improve algorithm. algorithm. The improved algorithm has the highest
Table 1.The accuracy rate of colorimetry experimental accuracy rate. When developing practical colorimetry
results system, this improved algorithm can be transplanted to
I II III VC platform, and this will absolutely improve the
The first 0.875 0.78125 0.953125 speed of processing. The work in this paper will offer
experiment the necessary foundation for the development of the
The second 0.828125 0.765625 0.96875 Computer-aided Colorimetric System for Baked
experiment Porcelain Tooth, which has certain clinical value.
The third 0.84375 0.78125 0.921875
experiment
Average 0.848958 0.776042 0.947917 Acknowledgement
accuracy rate
This work is completed with the help of Professor
Although the nine sub-areas subtract colorimetry (II) Desong Zhan who works in the oral cavity Hospital of
is more complex than the center area colorimetry (I), China Medical University, express our sincere thanks.
its accuracy is lower. The reason for that is the nine
sub-areas subtract colorimetry is just a simple References
calculation of the color character subtraction, the
interaction of the color character values in each area [1]K.E.Loshchilov, “Medical And Biological
have great influence on the final result. The differences Measurements--The Optikdent Dental CAD/CAM Suite”,
between the subtractions gotten from different models Measurement Techniques, Springer, New York, Vol. 49(12),
are very small in the two methods. That’s the reason for 2006, pp.1260-1263.
the low accuracy rate. Based on the above algorithms,
the improved algorithm is designed to compensate their [2] Liu Yang, “The Research Progress of Dental Colorimetric
deficiencies, whose accuracy rate is the highest. Technology”, Foreign Medical stomatology volumes,
As in Table 2, statistics of running time of the three Chengdu,Vol 33, No. 2, Mar. 2006,pp. 149-151.
algorithms are shown. Where, I stands for the center
[3]Shofu, ShadeEye-NCC, British Dental Journal,
area colorimetry; II stands for the nine sub-areas Vol.200(1),London, Jan. , 2006, pp.53.
subtract colorimetry; and III stands for the improve
algorithm. According to the experimental results, it is [4]Jiayu Zhang, Jiagen Ding, and Yonghong zeng,
known that center area colorimetry 's delay is the “Discussing on the mensuration of the porcelain tooth color
shortest of the three; the nine sub-areas subtract by HSB model”, China Agedness Stomatology Journal,
colorimetry has the lowest accuracy rate; the improved Beijing,Vol. 1(4), Dec.2003, pp.213-216.
algorithm has the highest accuracy rate , but the speed
is the slowest. [5]Zhaolin Sun, MATLAB6.X Image Processing, Tsinghua
University Press, Beijing, 2002, pp. 1-3.
Table 2.Statistics of delay of three algorithms
I II III [6]Tae-Woo KIM and Tae-Kyung CHO, “Teeth Image
The first experiment 0.220358 0.308892 0.250152 Recognition for Biometrics”, IEICE Trans. Inf. & Syst.,
(s) Vol.E89–D(3), Oxford University Press, Oxford ,MAR. 2006,
The second 0.209853 0.308800 0.332472 pp.1309-1313.
experiment
(s) [7]James Fondriest, “Shade Matching in Restorative
The third experiment 0.221339 0.240461 0.323906 Dentistry:The Science and Strategie”, The International
(s) Journal of Periodontics & Restorative Dentistry,
Average accuracy 0.217183 0.286051 0.302177 Quintessence Publishing Co., Berlin, 2003,Vol. 23(5),
rate (s) pp.467-479

131

Das könnte Ihnen auch gefallen