**6. Results**

The main performance evaluation criteria for the proposed algorithm are average execution time and LP character segmentation rate as shown in Table 1. The results clearly show an improvement when 2D Haar WT is used especially in terms of the character segmentation rate, which is tested on 6000 images combining both image sets of SD and HD. It is also noted that the execution time for character segmentation is close for both SD and HD images due to similar LP candidate size but higher character segmentation rate is observed at higher resolution.

The edges results from 2D Haar WT on an input LP candidate image and segmented characters are shown in figure 12 to figure 14.

The Haar edges are used as a reference without further processing of the Haar edges like thinning; we apply the edges comparison algorithm explained in Section 3.1 and compare location where an edge is verified if a match is found. The flow chart is shown in figure 10. The LP candidate has unique properties where the typical number of edges is between 100 to 2000 edges per plate. There are seven characters in UK LPs, a single character in a LP candidate contains between 30 to 150 edges, the gap of the character is between 2 to 4 pixels, the height of the character is about 20 pixels and width is about 16 pixels. This knowledge is applied to Connected Component Analysis (CCA) (Llorens, 2005) and a window (box) is drawn when a character is found. Finally, histogram analysis is applied to verify the

The proposed algorithms are optimized using similar experimental setup as reported in (Musoromy etal., 2010) and tested on Standard Definition (SD) and High Definition (HD) images that are a mixture of colour (day) and IR (night) with varying complexity levels such as over exposure, very dark and noisy. The proposed algorithm described in Section 4 forms a unified approach to resolve problems related to the above. The algorithm is implemented

A Windows host PC (2.4 GHz clock speed) with Code Composer Studio and a monitor

 A Texas Instrument's C64plus DSP (fixed-point DSP based on an enhanced version of the second generation high-performance, advanced Very-Long-Instruction-Word

Testing database of 5000 images of 768X288 resolutions (SD) and 1000 images of

The implementation of Haar WT based edge detector is performed using a TI's DSP. TI provides an image library which has a unique implementation of the DWT through a highly optimised image columns transformation, which provides horizontal and vertical wavelet transform functions (TI, 2006). We apply reconstruction to the vertical and horizontal

The main performance evaluation criteria for the proposed algorithm are average execution time and LP character segmentation rate as shown in Table 1. The results clearly show an improvement when 2D Haar WT is used especially in terms of the character segmentation rate, which is tested on 6000 images combining both image sets of SD and HD. It is also noted that the execution time for character segmentation is close for both SD and HD images due to similar LP candidate size but higher character segmentation rate is observed at

The edges results from 2D Haar WT on an input LP candidate image and segmented

(VLIW)) with minimum of 600MHZ clock speed and 1MB of RAM (TI, 2006) DSP host board with a JTAG interface debugger to provide interface between the DSP

1394X1040 resolutions (HD) provided by CitySync Ltd (CitySync, 2011)

presence of characters in a LP candidate.

**5. Experimental setup** 

in DSP using the following tools:

acting as baseline processor

and the host PC during debugging DSP algorithm

wavelet transform functions to obtain the edges.

characters are shown in figure 12 to figure 14.

**6. Results** 

higher resolution.

(a) (b)

Fig. 12. (a) The input LP Candidate (f(x,y)) and (b) the detected edges using 2D Haar WT (E(x,y))

Fig. 13. (a) The post- processed 2D Haar WT edges (EHaar) and (b) the detected edges in green (EFIN)

Fig. 14. (a) Character segmentation using histogram analysis (HA) and CCA, and (b) the segmented characters bounding box


Table 1. Algorithm profiling results

Real-Time DSP-Based License Plate

the original image

test results in Table 2.

**LP character segmentation algorithm** 

Using Haar WT

Using Haar WT

Character Segmentation Algorithm Using 2D Haar Wavelet Transform 19

 (a) (b)

 (c) (d)

 (e) (f) Fig. 16. The original license plate candidate of a higher resolution image (768x288) (a), one level (b), two levels (c), three levels (d), four levels (e) and five levels (f) decomposition of

The data set is partitioned further into day and night to provide more detailed analysis of

**HD (500 images)** 

Without Haar WT 89.2 93.4 90.1 95.4

(single level) 94.5 96.1 95.4 97.8

(two levels) 95.6 98.2 97.0 98.9

**SD (2500 images)** 

Table 2. Segmentation success rate for day and night images

**Day (3000 images) Night (Infra-Red)** 

**(3000 images)** 

**HD (500 images)** 

**SD (2500 images)** 

It is observed that when using high resolution images and reduced number of wavelet decomposition (small scale single level in our case) the result is noisier and more discontinuous edges while at lower resolution and high number of wavelet decomposition have an opposite effect. This was also reported by Qureshi (Qureshi, 2005). In our application, the former effect leads to failed character segmentation due to "bad edges" while the latter improve character segmentation rate at an expense of losing speed for real time application as shown in our results in Table 1. In this case, a good balance between image resolution and wavelet decomposition levels is required.

In conclusion, in Table 1, two levels provide better character segmentation rate compared to a single level. However, the slower times is the downfall, therefore we choose decomposition at a single level that meet real-time requirement, which also gives a good character segmentation rate.

The difference between lower and higher decomposition levels around the LP region are demonstrated in figure 15 for a lower resolution image and similarly, in figure 16 decomposition levels for higher resolution image are shown using similar post processing edge threshold. The results clearly shows images at higher resolution performs better at lower decomposition levels.

Fig. 15. The original license plate candidate of a lower resolution image 384x144 (a), one level (b), two levels (c), three levels (d), four levels (e) and five levels (f) decomposition of the original image

(e) (f)

It is observed that when using high resolution images and reduced number of wavelet decomposition (small scale single level in our case) the result is noisier and more discontinuous edges while at lower resolution and high number of wavelet decomposition have an opposite effect. This was also reported by Qureshi (Qureshi, 2005). In our application, the former effect leads to failed character segmentation due to "bad edges" while the latter improve character segmentation rate at an expense of losing speed for real time application as shown in our results in Table 1. In this case, a good balance between

In conclusion, in Table 1, two levels provide better character segmentation rate compared to a single level. However, the slower times is the downfall, therefore we choose decomposition at a single level that meet real-time requirement, which also gives a good

The difference between lower and higher decomposition levels around the LP region are demonstrated in figure 15 for a lower resolution image and similarly, in figure 16 decomposition levels for higher resolution image are shown using similar post processing edge threshold. The results clearly shows images at higher resolution performs better at

(a) (b)

 (c) (d)

(e) (f) Fig. 15. The original license plate candidate of a lower resolution image 384x144 (a), one level (b), two levels (c), three levels (d), four levels (e) and five levels (f) decomposition of

image resolution and wavelet decomposition levels is required.

character segmentation rate.

lower decomposition levels.

the original image

Fig. 16. The original license plate candidate of a higher resolution image (768x288) (a), one level (b), two levels (c), three levels (d), four levels (e) and five levels (f) decomposition of the original image

The data set is partitioned further into day and night to provide more detailed analysis of test results in Table 2.


Table 2. Segmentation success rate for day and night images

Real-Time DSP-Based License Plate

38–53. 10.1007/BF01456927.

Springer-Verlag, pp. 571–578.

*Vacuum, Xi'an, China*, pp. 780- 783.

*ItsApplications)*, Academic Press.

0-387-25280-3, New York, USA.

*Conference on* , vol., no., pp.92-97, 18-19 .

674–693.

Atlanta, USA.

Janeza, Croatia .

18.

17.

spru862a.

673.

Character Segmentation Algorithm Using 2D Haar Wavelet Transform 21

Desneux, P. & Legat J., D. (2000). A dedicated DSP architecture for discrete wavelet

Haar, A. (1911). Zur theorie der orthogonalen funktionensysteme, *Mathematische Annalen* 71:

Llorens, D., Marzal A., Palazon, V. & Vilar, J. M. (2005). Car License Plates Extraction and

Ma, X.D., Zhou C. & Kemp, I.J. (2000) "DSP based partial discharge characterization by

Mallat, S. (1999). *A Wavelet Tour of Signal Processing, Second Edition (Wavelet Analysis &* 

Mallat, S. (1989). A theory for multiresolution signal decomposition: the wavelet

Musoromy, Z., Bensaali F., Ramalingam S. & Pissanidis G. (2010). "Comparison of Real-Time

Palacios, G., Beltran, J. R & Lacuesta, R. (2011). Multiresolution Approaches for Edge

Patil, S. & Abel, E.W. (2006). Optimization of the Continuous Wavelet Transform for DSP

Paunwala, C.N., Patnaik, S. & Chaudhary, M. (2010). "An efficient skew detection of license

Qureshi, S. (2005). "*Embedded Image Processing on the TMS320C6000™ DSP*", Springer, ISBN

Roomi, S.M.M., Anitha, M., & Bhargavi, R. (2011). "Accurate license plate localization,"

TI, Texas Instruments. (2006). "*TMS320C64x+ DSP Cache User's Guide*", Literature number:

Wong, T.T., Leung, C.S., Heng, P.A. & Wang J. (2007). "Discrete wavelet transform

Recognition Based on Connected Components Analysis and HMM Decoding, *in Lecture Notes on Computer Science*, vol. 3522, J. S. Marques et al., Eds. New York:

wavelet analysis", *IEEE 19th Int. Symp. On Discharge and Electrical Insultaion in* 

representation, *IEEE Transactions on Pattern Analysis and Machine Intelligence* 11:

DSP-Based Edge Detection Techniques for License Plate Detection", *Sixth international Conference on Information Assurance and Security (IAS)*, pp 323-328,

Detection and Classification Based on Discrete Wavelet Transform, In: *Discrete Wavelet Transforms: Algorithms and Applications*, InTech, ISBN 978-953-307-482-5,

Processor Implementation, *Engineering in Medicine and Biology Society, 2005. IEEE-EMBS 2005. 27th Annual International Conference of the* , vol., no., pp.2787-2789, 17-

plate images based on wavelet transform and principal component analysis," *Signal and Image Processing (ICSIP), 2010 International Conference on* , vol., no., pp.17-22, 15-

*Computer, communication and Electrical Technology (ICCCET), 2011 International* 

on consumer-level graphics hardware*", IEEE Trans. Multimedia* 9 (3) 668–

transform. *Integr. Comput.-Aided Eng.* 7, 2 (April 2000), 135-153.

It is also noted in Table 2 that there is a small character segmentation success advantage in images taken at night compared to images taken in the day time. This can be explained due to the fact that at night, an Infra-Red (IR) camera is used to capture license plate which provides good images due to license plate's reflectivity to IR camera where the other objects in the background are not captured.

As well as the "bad edges", there are a number of factors that cause license plate character segmentation failure including;

