Volume 1, No. 7, September 2012 ISSN 2278-1080 The International Journal of Computer Science & Applications (TIJCSA) RESEARCH PAPER Available Online at http://www.journalofcomputerscience.com/ Edge Detection for Dental X-ray Image Segmentation using Neural Network approach Dr. N. Senthilkumaran Department of Computer Science and Applications Gandhigram Rural Institute Deemed University Gandhigram, India nsenthilkumaran@hotmail.com Abstract In the Digital Image Processing, edge detection playing a vital role. An edge is the boundary between an object and the background, and indicates the boundary between overlapping objects. Artificial Neural Network, inspired by the way of biological nervous systems such as human brains process information and it is an information processing system which contains a large number of highly interconnected processing neurons. These neurons work together in a distributed manner to learn from the input information, to coordinate internal processing, and to optimize its final output. In this paper, the main aim is to study the theory of edge detection for dental x-ray image segmentation using a neural network approach. Keywords- edge detection, image segmentation, neural network, dental x-ray image. 1. Introduction Edge detection is an important but rather difficult task in image processing and analysis[1]. Edges in images are the curves that characterize the boundaries of objects[2]. Edges contain important information of objects such as shapes and locations, and are often used to distinguish different objects and/or separate them from the background in a scene[3]. It can significantly reduce the amount of data to be processed in the subsequent steps such as feature extraction, image segmentation, registration, and interpretation[4]. Edge detection has found many applications in pattern recognition, image analysis, and computer vision. Since edges are associated with abrupt intensity changes, edge detection is the process to identify and locate such sharp intensity contrasts in an image and it is well known that slow changes correspond to small values of derivatives while fast changes correspond to large values of derivatives[5]. Based on two-dimensional spatial filter or gradient operator is often employed in conventional edge detection algorithms. This filter is designed to be sensitive to detect the gradient of image intensity while yields no response to non-edge regions or the areas with constant intensity in the image[6]. A variety of filters or masks have been developed to detect various types of edges. Once the mask is constructed, it convolves with the entire image, pixel by pixel, to detect edges[6][7]. Recently, neural networks have been applied to edge detection based on the adaptive learning ability and nonlinear mapping ability; neural networks can be trained to detect edges and can serve as nonlinear filters once they are fully 2012, http://www.journalofcomputerscience.com - TIJCSA All Rights Reserved 8
trained[5][8]. The multi-layer feed-forward neural networks for the edge detection of the Dental X-ray image of a bridge. Multiple neural networks are trained by synthetic edge patterns; each one of them can detect a specific edge pattern. If desired, one can also combine the outputs from the group of neural networks to detect multiple types of edges in images[4]. The paper is organized as follows. In section 2 discuss about overview of edge detection techniques. The neural network approach for this research is discussed in section 3. In section 4, details about dental x-ray image segmentation. In section 5, computer simulation results are presented. Section 6 concludes this paper and also gives the direction for future works. 2. Overview of Edge Detection Techniques Edge detection is a process that detects the presence and location of edges constituted by sharp changes in color, intensity of an image[3][10]. Edge is a part of an image that contains significant variation[2]. The edges provide important visual information since they correspond to major physical, photometrical or geometrical variations in scene object[1]. Physical edges are produced by variation in the reflectance, illumination, orientation, and depth of scene surfaces[6]. Since, it can be proven that the discontinuities in image brightness are likely corresponding to discontinuities in depth, discontinuities in surface orientation, changes in material properties and variations in scene illumination[4]. In this case, the result of applying an edge detector to an image may lead to a set of connected curves that indicate the boundaries of objects, the boundaries of surface markings as well curves that correspond to discontinuities in surface orientation[2]. So, it is not always possible to obtain such ideal edges from real life images of moderate complexity[3]. Since image intensity is often proportional to scene radiance, physical edges are represented by changes in the intensity function of an image[1]. Therefore, it should be mandatory to find out the occurrence in perpendicular to an edge[6][9]. 3. Neural Network Approach Digital images are segmented by using neural networks in two step process[3]. First step is pixel classification that depends on the value of the pixel which is part of a segment or not[4]. Second step is edge detection that is the detection of all pixels on the borders between different homogeneous areas which is part of the edge or not[6]. Several neural networks are available in the literature for edge detection[5][10]. A multi-layer feed-forward artificial neural network (ANN) model for edge detection is well known that ANN can learn the inputoutput mapping of a system through an iterative training and learning process, thus ANN is an ideal candidate for pattern recognition and data analysis[2][9]. The ANN model employed in this research has one input layer, one output layer, and one hidden layer[5]. The neural network model is a multi-input, single-output system[4]. The output neuron is linear, the activation function for each hidden neuron is chosen as the hyperbolic function[6]. The weights of the neural network are updated using the Levenberg-Marquardt algorithm[3] to minimize the following objective function, 2012, http://www.journalofcomputerscience.com - TIJCSA All Rights Reserved 9
where d is the desired output and y is the output of neural network, e is the output error (i.e., the difference between the neural network output and the desired output), k is the index of a training pair[4]. Let W be the weight matrix of the neural network, then where Ja is the first order derivative of the error function with respect to the neural network weight, e is the output error (i.e., the difference between the neural network output and the desired output), µ is a learning parameter[5]. Total 22 pixels were used in the ANN structure[6]. These pixels are arranged in the form of 9 pixels for input layer, 12 pixels for hidden layer and 1 pixel for output layer[4]. ANN uses feed-forward propagation neural network architecture[3]. Fig.1. Aritificial Neural Network Architecture for image edge detection Input layer Training is obtained 3x3 image mask taken from original image[4][10]. And then, they are used in the input of ANN for training[5]. Here, each of 9 cells at input layer connects to each of 12 cells at hidden layer[3]. Values which enter from 9 input cells are calculated with a gradient method and results are used at hidden layer input[7]. Hidden layer and output layer used for a BP method[6]. The error of network output is distributed to connected weights and then parameters are iteratively updated to minimize the error[4]. Network training and this operation are repeated. Practically, edge detection operation completes using mask is made with a 3x3 matrix around the image[6]. Matrix values are multiplied with picture s pixels, and the results are added to other result values, then the new value is written to result on image[5]. After this operation, edges are defined on picture. Masked matrix values are exchanged according to the method of edge detection[3]. 4. Dental X-ray Image Segmentation Dental X-rays are pictures of the teeth, bones, and soft tissues around them to help find problems with the teeth, mouth, and jaw[11]. Dental X-ray pictures can show cavities, hidden dental structures and bone loss that cannot be seen during a visual examination[11]. Dental X-rays are detecting problems in the mouth such as tooth decay, damage to the bones supporting the teeth, and dental injuries (such as broken tooth roots)[11]. Dental X-Rays are often done to detect these problems early, before any symptoms are experienced and detect 2012, http://www.journalofcomputerscience.com - TIJCSA All Rights Reserved 10
teeth that are abnormally placed or don't break through the gums properly[11]. Teeth that are too crowded to break through the gums are called impacted[11]. Detect cysts, solid growths (tumors), or abscesses caused by certain dental problems[11]. Evaluate the presence and location of permanent teeth growing in the jaw in children who still have their primary teeth[11]. Without Dental X-Rays, dentists would miss the early stages of decay between teeth[11]. (a) (b) (c) Fig.2. Dental X-ray images, (a) Original Image, (b) and (c) Dental X-ray images 5. Results 2012, http://www.journalofcomputerscience.com - TIJCSA All Rights Reserved 11
(a) (b) (c) (d) (e) (f) Fig. 3. Dental X-ray Images, (a) original images, (b) using Roberts operator, (c) using Prewitt operator, (d) using Sobel operator, (e) using canny operator, (f) using Artificial Neural Network approach. 6. Conclusion This paper presents a method of edge detection using neural network. Dental X-ray image is used for edge detection to produce the edges of the image. Neural network uses these edges to learn edges of all images. It has been shown that the neural network was provided to high performance in this application. Especially, neural network has flexible edge detection on shady segments. Find problems in the mouth such as tooth decay, damage to the bones 2012, http://www.journalofcomputerscience.com - TIJCSA All Rights Reserved 12
supporting the teeth, and dental injuries. Dental X-rays are often done to find these problems early, before any symptoms are present. Find teeth that are not in the right place or do not break through the gum properly. Find cysts, solid growths, or abscesses. This work may be extended for plan treatment for large or extensive cavities, root canal surgery, placement of dental implants, and difficult tooth removals. In this paper, studied neural network technique and investigated the features and the usefulness and applied to edge detection of dental x-ray images and have confirmed its efficiency. References [1] L. Canny, A computational approach to edge detection, IEEE Trans. on Pattern Analysis and Machine Intelligence, vol. 8 no. 1, pp. 679-698, 1986. [2] D. Marr, E. Hildreth, Theory of edge detection, Proc. of Royal Society Landon, B(207): 187-217, 1980. [3] A. J. Pinho, L. B. Almeida, Edge detection filters based on artificial neural networks, Pro. of ICIAP'95, IEEE Computer Society Press, p.159-164, 1995. [4] Wang, X.Q. Ye, W.K. Gu, Training a NeuralNetwork for Moment Based Image Edge Detection Journal of Zhejiang University SCIENCE(ISSN 1009-3095, Monthly), Vol.1, No.4, pp. 398-401 CHINA, 2000. [5] Y. Xueli, Neural Network and Example, Learning, Press of Railway of China, Beijing, 1996. [6] H. Abdi, D. Valentin and B. Edelman, Neural networks, Thousand Oaks (CA), Sage, 1998. [7] N. Senthilkumaran and R. Rajesh, Edge Detection Techniques for Image Segmentation A Survey of Soft Computing Approaches, International Journal of Recent Trends in Engineering (Computer Science), Academy Publisher, Finland, Vol. 1, No.2, ISSN. 1797-9617, May 2009, pp. 250-254. [8] N. Senthilkumaran and R. Rajesh, A Study on Edge Detection Methods for Image Segmentation, Proceedings of the International Conference on Mathematics and Computer Science, (ICMCS-2009), organized by Loyola College, Chennai, ISBN: 978-81-8371-194-4, 2009, Vol. I, pp. 255-259. [9] N. Senthilkumaran and R. Rajesh, Edge Detection Techniques for Image Segmentation A Survey, Proceedings of International Conference on Managing Next Generation Software Applications, (MNGSA-08), organized by Department of Computer Applications, Karunya University, Coimbatore, ISBN: 978-81-907870-0-0, 2008, pp.749-760. [10] N. Senthilkumaran and J. Suguna, Neural Network Technique for Lossless Image Compression using X-Ray Images, International Journal of Computer and Electrical Engineering, Vol.3, No.1, ISSN : 1793-8163, February 2011, pp. 17-23. [11] A. K. Jain and H. Chen, Matching of dental x-ray images for human identification, Pattern Recognition, 37:1519 1532, 2004. 2012, http://www.journalofcomputerscience.com - TIJCSA All Rights Reserved 13