Circuits and Systems, 2011, 2, 320-325
doi:10.4236/cs.2011.24044 Published Online October 2011 (
Copyright © 2011 SciRes. CS
An Efficient Method for Vehicle License Plate Detection in
Complex Scenes
Mahmood Ashoori-Lalimi, Sedigheh Ghofrani*
Electrical Engineering Department, Islamic Az ad University, South Tehran Branch, Tehran, Ira n
E-mail:, *
Received April 2, 2011; revised Au gu st 23, 2011; accepted August 30, 2011
In this paper, we propose an efficient method for license plate localization in the images with various situa-
tions and complex background. At the first, in order to reduce problems such as low quality and low contrast
in the vehicle images, image contrast is enhanced by the two different methods and the best for following is
selected. At the second part, vertical edges of the enhanced image are extracted by sobel mask. Then the
most of the noise and background edges are removed by an effective algorithm. The output of this stage is
given to a morphological filtering to extract the candidate regions and finally we use several geometrical
features such as area of the regions, aspect ratio and edge density to eliminate the non-plate regions and
segment the plate from the input car image. This method is performed on some real images that have been
captured at the different imaging conditions. The appropriate experimental results show that our proposed
method is nearly independent to environmental conditions such as lightening, camera angles and camera dis-
tance from the automobile, and license plate rotation.
Keywords: License Plate Detection, Image Enhancement, Background and Noise Removing, Morphological
1. Introduction
With the rapid development of highway and the wide use
of vehicle, researchers start to pay more attention on ef-
ficient and accurate intelligent transportation systems
(ITS). It is widely used for detecting car’s speed, security
control in restricted areas, highway surveillance and
electric toll collection [1]. Vehicle license plate (VLP)
recognition is one of the most important requirements of
an ITS. Although any ITS and specifically any VLP rec-
ognition contains two part in general, license plate detec-
tion and recognition, detecting and segmenting VLP
correctly is most important because of existing condi-
tions such as poor illumination, vehicle motion, view-
point and distance changes. The problem of automatic
VLP recognition has been studied since 1990s. The first
approach was based on characteristics of boundaries
[2,3]. In this method, an image was binarized and then
processed by certain algorithms, such as Hough trans-
form, to detect lines. In general, the most common ap-
proaches for VLP detection include texture [1,4], color
feature [5], edge extraction [6], combining edge and color
[7], morphological operation [5,8] and learning-based
method [9]. Using color feature is benefit when lighten-
ing is unchanged and stable. However methods based on
edge and texture are nearly invariant to different illumi-
nation and so they are widely used for VLP detection.
These methods use the fact that there are many charac-
ters in the license plate, so the area contains rich edge
and texture information. Zhang et al. [9] proposed learn-
ing-based method using AdaBoost for VLP detection.
They used both global (statistical) and local (Haar-like)
features to detect the license plate.
In this paper, we do pre processing for image en-
hancement at first. The some regions are candidate as a
license plate during three procedures. Finally considering
geometrical features, the license plate is segmented
nearly independent of image capturing conditions .
This paper is organized as follows: in Section 2, dif-
ferent styles of Iranian license plates are illustrated in
Section 3. We express how our image bank is provided.
In Section 4, the proposed algorithm is described and in
Section 5 the experimental results are reported. Finally,
in Section 6 we have conclusion.
2. Iranian Vehicle License Plate
We have considered 3 classes for Iranian VLP, they are
private, public (such as taxi, truck and bus) and govern-
mental vehicles. Each class has own plate and character
color. In addition, though Farsi characters are 32, only
some characters are used for VLP. Color arrangement,
characters and outline of the Iranian VLP are shown in
Table 1.
3. Provided Image Bank
As respects, the aim of this paper is detecting the Iranian
license plates in images with complex sc enes. Du e to the
unavailability of required images, in several stages by
using 2 digital cameras and mobile cameras, we have
provided 350 images under various illumination (light-
ening), different distances and angles of stationary and
moving vehicles. After providing images, in order to
increase the processing speed and facilitate the license
plate detection, input color image is converted to gray-
scale image. The size of images is 640 × 480 pixels.
4. Efficient License Plate Detection
Our proposed method is composed of several parts, Fig-
ure 1 shows the flowchart.
4.1. Pre Processing
Low contrast may have the most important effect on
failing a license plate detection algorithm. Severe light-
ening conditions, changing plate orientation and various
distances are main reasons for having low contrast and
quality in the car images. Therefore, contrast enhance-
ment seems to be necessary, specially at locations where
might be a license plate. So, in following we improve
different images using two methods, they are intensity
variance [6] and edge density [7], and choose the best for
pre processing images.
4.1.1. Intensity Variance
Zheng et al. [6] used the local variance of pixel intensi-
ties to improve image contrast at regions that may be
plate. They proposed an enhancement function which
increases image contrast at regions that local variance of
intensity is around 20. The enhancement function was
suggested as follows:
ijij ij
ijwij ww
where ij
, and ij
denote the intensities of the pixel in
the input grayscale image and enhanced image, and
is a window centered on pixels of grayscale image. ij
and ij
are average luminance and standard deviation
respectively. The enhanced coefficient is defined as fol-
3if 020
220 1
400 3
()if20 60
220 1
ij ij
With respect to Figure 2, the intensities of pixels in
the input grayscale images with local variance between 0
and 60 are enhanced.
Figure 3 shows the result of image enhancement using
the zheng’s method.
4.1.2. Edge Density
Abolghasemi et al. [7] used the density of vertical edges
(instead of the variance of intensity) as criterion for lo cal
enhancement of car image. License plate of the car con-
sist of several characters (8 characters for Iranian VLP),
so the license plate area contains rich edge information.
We can employ the edge information to find the location
of plate in an image. At first, they [7] used the vertical
sobel mask and obtained the gradient image:
Table 1. Different styles of Iranian VLP.
Vehicle Type Plate Color Character ColorOutline of VLP
Private (automobil e) white black
Public (taxi, truck and bus) yellow black
Governmental red white
Copyright © 2011 SciRes. CS
Figure 1. Flowchart of our proposed system for VLP detec-
020 406080100 120
Enhancement Coefficient
Local Standard Deviation
Figure 2. The graph of enhancement coefficient, ()
based on the local standard devi ati o n, i
Figure 3. (a) input grayscale image. Iranian license plate
before (b) and after (c) enhancement using intensity vari-
ance method.
Then, they compared pixel values with a predefined
threshold and the vertical edge image has been achieved.
In the next step, the vertical edge image is convolved
with the 2-D Gaussian kernel and estimation of the edge
density is yielded. The results on a sample image are
shown in Figure 4.
In order to enhance the input image with respect to the
estimations of edge density, an enhancement coefficient
is suggested as follo ws:
ijij ij
ijwij ww
fIII (4)
where ,
and ij
are explained in the previous
fij is the weighting function, regarding the
estimation of edge density. This function is sketched in
Figure 5.
As can be seen in Figure 5, the intensity of pixels with
the edge density among 0.15 to 0.45 is to be enhanced.
The enhancement coefficient
f is defined as fol-
3, if00.15
20.15 1
0.15 3,if 0.150.5
20.15 1
0.5 0.15
1, if0.5
Figure 6 shows the result of enhancement by this me-
Even though for normal images, as it can be seen in
Figure 7(a), both intensity variance [6] and edge den sity
[7] can improve the VLP, when the distance and angle
(b) (c)
Figure 4. (a) Input grayscale image, (b) ve rtical edge image
and (c) the edge-density estimation image.
Copyright © 2011 SciRes. CS
00.2 0.4 0.6 0.81
Enhancement Coefficient
Normalized Edge Density
Figure 5. The graph of enhancement coefficient,
based on normalized edge densit y, ij
Figure 6. (a) input grayscale image. Iranian license plate
before (b) and after (c) enhancement using edge density
Figure 7. (a), (b) input gray scale image . Iranian l icense pl ate
(c),(f) without enhancement (d),(g) improved using edge
density method (e), (h) improved using intensity variance
between camera and vehicle are increased, zheng’s
method fails while abolghasemi’s method already im-
proves the quality of VLP con siderably, Figure 7(b). so,
in this work we employ edge density method for pre
4.2. Detecting the VLP
After enhancing an input imag e by using suitab le method
(edge density), we should detect any existed license plate
in the improved image. We do the following stages for
this purpose.
4.2.1. Vertical Edge Detection
Edge detection is one of the most important processes in
image analysis. An edge represents the boundary of an
object which can be used to identify the shapes and area
of the particular object. When there is contrast difference
between the object and the background, after applying
edge detection, the object edges will be illustrated. We
select the vertical sobel operator, Equation (3), to detect
the vertical edges.
After convolving the enhanced car image with the ver-
tical sobel operator, an estimation of vertical gradient
image is yielded. Finally, we get a binary image, as
shown in Figure 8(a), by using a thres ho ld value.
4.2.2. Filter-O ut the L on g a nd Sh or t Ed ges
After extracting vertical edges from the enhanced image,
using morphological filtering obtains candidate regions
those may be a license plate. But, as it can be seen in
Figure 8(a), there are many long background and short
noise edges that may interference in the morphological
filtering process. In order to resolve this problem, an
effective algorithm is used to remove the background
and noise edges [6]. The filter-out image after removing
unwanted edges is shown in the Fi gur e 8 (b).
4.2.3. Candidate Regions Extraction Using
Morphological Filtering
Morphological filtering is used as a tool for extracting
image components and so representing and describing
region shapes such as boundaries. In this part, we use a
morphological operation for extracting candidate regions.
Hence, we implement the morphological closing and
opening that defined as follows:
Closing oper ation ()
 
Opening operation ()
mnmn mn
 
, and
denote dilation and erosion opera-
tions, respectively. mn
denote a structuring element
with size mn
, all entries in mn are one. The output
of this stage is shown in Figure 9(a).
4.2.4. Accura t e Loc ati o n of L i cense Plat e
After using morphological filtering, still many regions
(a) (b)
Figure 8. (a) vertical e dge image, (b) removing background
and noise edges (filter-out the long and shor t edges).
Copyright © 2011 SciRes. CS
(a) (b)
Figure 9. (a) Connected regions obtained from morpho-
logical process, (b) after applying features (c) cropped im-
are candidating as a license plate. So we consider some
features such as area, aspect ratio (height per width) and
edge density in order to discard wrong candidate regions.
Values for these features are set experimentally based on
our test images. These features are scale-, luminance-
and rotation-variant. Progressive of using these features
to remove non-plate candidate regions can be seen in
Figure 9(b) .
5. Experimental Results
We have run our proposed algorithm on laptop Core 2
Due CPU 2.26 MHz with 2 GB of RAM under MAT-
LAB R2008b environment. In Section 3, we described
how the vehicle images are provided. Some sample im-
ages of our database are shown in Figure 10.
Now, in order to evaluate the accuracy of our pro-
posed method, we categorize the provided database into
three categories including: Angle (high angle (>30 de-
gree & <60 degree) and low angle (<30 degree)), Dis-
tance (short distance (<4 m) and normal distance (>4 m
& <12 m) and low quality or contrast (evening, sunlight,
Figure 10. Some sample images of our database.
and rainy or cloudy weather).
Table 2 shows the accuracy of the proposed algorithm
under mentioned co nditions.
As it is written in Table 2, the average accuracy
achieved by our proposed method for license plate detec-
tion is 95.2%. It means that we could detect 333 Iranian
license plate correctly. Although our achieved accuracy
is less than what zheng [6] got but we believe our pro-
vided image bank is more complex. Because of showing
in Figure 7, the zheng’s method is sensitive to real situa-
tions such as existing long distance and high angle. Ta-
ble 3 shows the result of license plate detection for sev-
eral vehicle images in our categories.
Table 2. Accuracy achieved by our proposed method.
Images in Different
Situations Number of
Images Accuracy (%)
short 100 97
Distance normal 100 95
low 50 98
Angle high 50 94
Low Quality 50 92
Average (350 Images) 95.2
Table 3. Detected license plate by our proposed algorithm.
Input Gray
Scale Image Detected
License Plate Different
short distance
normal dis-
low angle
high angle
low quality
Copyright © 2011 SciRes. CS
Copyright © 2011 SciRes. CS
6. Conclusions
In this paper, an efficient license plate detection method
is proposed which performed on images with complex
scenes. We used edge density as pre processing for im-
age enhancement. Then by using vertical sobel mask,
removing background and noise edges, and employing
morphological filter, some regions are candidate as li-
cense plate. Finally considering geometrical features
(such as area of the regions, aspect ratio and edge den-
sity), the license plate from the input car image is ex-
tracted. Although the proposed algorithm performs on
the Iranian vehicle license plates under various situations
such as different lightening conditions, varied distances
and existence angle between the camera and the vehicle
and varied weather conditions, we believe its perform-
ance is yet appropriate if we try to detect the foreign li-
cense plate.
7. References
[1] V. Shapiro, G. Gluhchev and D. Dimov, “Toward a Mul-
tinational Car License Plate Recognition System,” Ma-
chine Vision Application, Vol. 17, No. 3, July 2006, pp.
173-183. doi:10.1007/s00138-006-0023-5
[2] V. Kamat and S. Ganesan, “An Efficient Implementation
of the Hough Transform for Detecting Vehicle License
Plates Using DSP’s,” Proceedings of Real Time Tech-
nology and Applications, Chicago, 15-17 May 1995, pp.
58-59. doi:10.1109/RTTAS.1995.516201
[3] Y. Yanamura, M. Goto and D. Nishiyama, “Extraction
and Tracking of the License Plate Using Hough Trans-
form and Voted Block Matching,” IEEE proceedings of
Intelligent Vehicles Symposium, Columbus, 9-11 June
2003, pp. 243-246.
[4] C. N. E. Anagnostopoulos and I. E. Anagnostopoulos, “A
License Plate Recognition Algorithm for Intelligent
Transportation System Applications,” IEEE Transaction
on Intelligent Transportation System, Vol. 7, No. 3, 2006,
pp. 377-392.
[5] W. G. Zhu, G. J. Huo and X. Jia, “A Study of Locating
Vehicle License Plate Based on Color Feature and
Mathematical Morphology,” 6th International Confer-
ence on Signal Processing, Vol. 1, 2002, pp. 748-751.
[6] D. Zheng, Y. Zhao and J. Wang, “An Efficient Method of
License Plate Location,” Pattern Recognition Letters, Vol.
25, No. 6, 2005, pp. 2431-2438.
[7] V. Abolghasemi and A. R. Ahmadyfard, “An Edge-Based
Color-Aided Method for License Plate Detection,” El-
sevier Journal of Image and Vision Computing, Vol. 27
No. 8, 2009, pp. 1134-1142.
[8] J. W. Hsieh, S. H. Yu and Y. S. Chen, “Morphology-
Based License Plate Detection from Complex Scenes,”
16 th International Conference on Pattern Recognition,
Vol. 3, 2002, pp. 176-179.
[9] H. Zhang, W. Jia, X. He and Q. Wu, “Learning-Based
License Plate Detection Using Global and Local Fea-
tures,” ICPR, Vol. 2, 2006, pp. 1102-1105.