Conglomeration for color image segmentation of Otsu method, median and Adaptive median Puneet Ranout 1, Anubhooti Papola 2 and Devesh Mishra 3 1 PG Student, Department of computer science and engineering, Faculty of Technology, Uttarakhand Technical University, Dehradun-248007 2 Astt. Professor, Department of computer science and engineering, Faculty of Technology, Uttarakhand Technical University, Dehradun-248007 3 Astt. Professor, Department of computer science and engineering, Faculty of Technology, Uttarakhand Technical University, Dehradun-248007 ABSTRACT The objective of this work is to develop efficient method for color image segmentation. In the first part of the work we have done the thresholding of the image. The Otsu method is used for the thresholding of image. The traditional Otsu method for gray channel image segmentation were applied for each of the R,G, and B channels separately to determine the suitable automatic threshold for each channel. After that, the new modified channels are integrated again to formulate a new color image. The resulted image suffers from some kind of distortion. To avoid this, in the second part the resulted image is passed through median and then the resulted image from the median is passed through the Adaptive median. The smoothened image from the Adaptive median is the resulted image. Experimental results were presented on a variety of test images. Keywords: Thresholding,, Adaptive median, Otsu Method, color image segmentation. 1. INTRODUCTION Image segmentation partitions an image into non-overlapping regions. A region is defined as a homogenous group of connected pixels with respect to a chosen property [1]. The segmentation in color image analysis is one of the most important problems. The fundamental idea in color image segmentation is to consider color uniformity as a relevant criterion to partition an image into significant regions [2]. Image segmentation results have an effect on image analysis and it following higher order tasks. Image analysis includes object description and representation, feature measurement. Higher order task follows classification of object.. Hence characterization, visualization of region of interest in any image, delineation plays an important role in image segmentation [3]. In a grayscale image, the difference between two pixels can simply be measured as the difference in brightness between these two pixels, because that is the definition of a difference between two shades of gray. For color images, only considering the brightness is not enough, since two distinct colors can be of the same brightness [4]. Usually Segmentation is the first task for any image processing. The pursuant tasks depend on the nature of segmentation. For this reason, an adequate attention is taken to improve the quality of segmentation [5]. In the gray level image segmentation a problem is mostly occurs when an image has a varying gray level background. One of these problem is when the image contains wide range of gray levels or progressively varying shadows. This is the color intensity problem in the gray scale images. In the high details image,the detection procedure for human can be done for one or two dozen intensity levels at any point due to luminosity accommodation [6]. Human eye can distinguish thousand of color shadows and intensities. The gray scale segmentation techniques like histogram thresholding, neural networks, fuzzy methods have been extended for color image segmentation by using RGB, CYM, HSI color space system etc. has demonstrated an effective technique to suppress impulse noise while preserving signal changes [7]. is a non-linear use to smooth image [8]. one disadvantage of median is high blurring of image when window size is large [9]. aims to change noisy pixels in such way to look like its nearby neighbors [10]. The work has been implemented using MATLAB R2013b. The paper is structured as follows: section (II) tells about the Related work, Section (III) deals with Experimental results, Finally the conclusion section may be seen in section (IV). Volume 3, Issue 9, September 2014 Page 307
2. RELATED WORK In this technique we will find the smoothening of the image and will check image for different window sizes. The different window size shows a different output. In this the edges of the image will be smooth as compare to the older algorithm of the segmentation for median. In the proposed algorithm firstly an image is uploaded for segmentation and the channels R,G, B of image are separated. Then the Otsu automatic thresholding method is applied in each channel for thresholding the image. After thresholding each channel, the thresholded output of all the channels are combined together to form a new image. the new image formed is a colored image. Moreover median ing technique could be applied to smooth the image. The R,G,B channel of the new image are separated and then the median is applied on each channel. The output of each channel is combined. The output image of median is then passed to Adaptive median to get more smoothened. From the median output image the channels R,G,B are separated again and the adaptive median flter is applied on them. The output from each channel the combined. The noticeable result may be cleared form human eye, as the edges are smoothed and the blurring is less as compare to previous work. As in this paper main contribution is that, when increasing the window size noticeable blurring results may be clear from human eye. Hence a suitable window size must be determined carefully is such a way that keeps the blurring amount in the safety side away from the distortion and high blurring. The suitable window size that was obtained by the proposed algorithm was found to be (15 15). (x, y) = Otsu Adaptive (x,y) Otsu (x,y ) (x,y ) Adaptive median (x,y) Otsu 1(x,y) + Adaptive Fig no. 1 Data flow Diagram The above written equation shows the (x, y ) is an image in which x and y are coordinates of image. In the other part of the equation represents the image but here. i is the no. of channels (R,G,B) of image. The value of i=1,2,3, so and i=1 means red channel of the image. i=2 means the green channel of the image and i=3 means the blue channel of the image. Acc to figure(2a), it can be seen that the implementation of traditional Otsu method for each channel of the R,G,B channels will produce some kind of noise regions. Therefore to make these regions smoother, an adaptive median could be applied with K K window size to get rid of these noisy regions. This process is very useful in object recognition, consequently, image segmentation. It must be mentioned that whenever there is an increase in the block size, there will be increase in the smoothness process. Hence four types of block sizes have been applied which are: 3 3, 7 7, 11 11,15 15 (2b) is high segmented than the traditional Otsu method (2a). Experimental results are presented concerning the previously proposed method. Several (512 512) test images have been used to implement the proposed method for color image segmentation (figure 3). The experimental results have been shown in figures (4),(5),(6) (7). Furthermore a combination of window sizes have been applied which are (3 3,7 7,11 11,15 15) in case of adaptive median ing process. The difference between the median and the adaptive median can be seen easily in the experimental results. In the adaptive median the color intensity and color distribution over Volume 3, Issue 9, September 2014 Page 308
the image could be better than the median. Hence from the ocular results it is very clear that the convenient window size is (15 15) for all the test images that have been used. Fig no.2 (a) original image (b) Traditional Otsu image (c) Hybridization between Ostu for each of R,G,B channels method and 7 7 3. EXPERIMENTAL RESULTS In this Section, details of the implementation 3 3 Fig no.3 Variety of (512512) test images (Baboon, Lena, Pepper and Airplane) Fig no. 4 Fig no. 5 Fig no. 6 fig no. 7 Volume 3, Issue 9, September 2014 Page 309
Adaptive median 3 3 3 3 Fig no. 8 Fig no. 9 Adaptive 3 3 Fig no. 10 Fig no. 11 7 7 Fig no. 12 Fig no. 13 Fig no. 14 Fig no. 15 Volume 3, Issue 9, September 2014 Page 310
Adaptive median 7 7 7 7 Fig no. 16 Fig no. 17 11 11 Fig no. 18 Fig no. 19 Adaptive median 11 11 Fig no. 20 Fig no. 21 Fig no. 22 Fig no. 23 Volume 3, Issue 9, September 2014 Page 311
11 11 Fig no. 24 Fig no. 25 Adaptive median 11 11 15 15 Fig no. 26 Fig no. 27 Fig no. 28 Fig no. 29 Adaptive 15 15 Fig no. 30 Fig no. 31 Volume 3, Issue 9, September 2014 Page 312
15 15 Adaptive 15 15 Fig no. 32 Fig no. 33 4. CONCLUSIONS Fig no. 34 Fig no. 35 In this paper, a new approach for color image segmentation has been presented which is based on the conglomeration of otsu method, median and adaptive median. The implementation of otsu method to the R,G,B channels with median ing will show some kind of noise and to get rid of this noise an adaptive median ing process was proposed. The edges will be smooth and the intensity of color would be better than the median ing output image. The main conclusion comes here is that the increase in window size (K K) that was implemented in ing process will increase the interior homogeneity of the regions and objects inside image. Hence in this paper, a window size 15 15 shows a good relevant result. This method is easy to implement and has a high rapidity. According to the result in the previous section the proposed technique is recommended in Biometric recognition, medical image processing, in agriculture (inspections of fruits and vegetables). In this the future work could be checking the working of otsu method with other that can help in better image quality. REFERENCES [1] FirasAjilJassim, Fawzi H. Altaani Hybridization of Otsu Method and Filter for Color Image Segmentation, International Journal of Soft Computing and Engineering (IJSCE) ISSN: 2231-2307, Volume-3, Issue-2, May 2013. [2] L. Busin, N. Vandenbroucke, and L. Macaire, Color spaces and image segmentation, Advances in Imaging and Electron Physics, vol. 151, 2008, pp. 65-168. [3] H.P. Narkhede, Review of Image Segmentation Techniques, International Journal of Science and Modern Engineering (IJISME) ISSN: 2319-6386, Volume-1, Issue-8, July 2013. [4] K. K. Singh, A. Singh, A Study of Image Segmentation Algorithms for Different Types of Images, International Journal of Computer Science Issues, Vol. 7, Issue 5, 2010. [5] R. C. Gonzalez and R. E. Woods. Digital Image Processing, Prentice Hall, New Jersey 07458, second edition, 2001. [6] N. Ikonomakis, K. N. Plataniotis, A. N. Venetsanopoulos, Color Image Segmentation for Multimedia Applications, Journal of Intelligent and Robotic Systems, vol.28, 2000, pp. 5 20. [7] J. M. C. Geoffrine and N. Kumarasabapathy, Study And Analysis Of Impulse Noise Reduction Filters, Signal & Image Processing : An International Journal (SIPIJ), Vol.2, No.1, March 2011. Volume 3, Issue 9, September 2014 Page 313
[8] Computer Vision CITS4240 School of Computer Science & Software Engineering, The University of Western Australia. [9] H. Gomez-Moreno, S. Maldonado-Bascon, F. Lopez-Ferreras, M. Utrillamanso And P. Gil-Jimenez, A Modified Filter for the Removal of Impulse Noise Based on the Support Vector Machines, Proceedings of the 7th International Work-Conference on Artificial and Natural Neural Networks: Part II: Artificial Neural Nets Problem Solving Methods (IWANN 03), Menorca, Spain, 2003, pp. 536-543. [10] A. A. Gulhane, A. S. Alvi, Noise Reduction of an Image by using Function Approximation Techniques, International Journal of Soft Computing and Engineering (IJSCE), vol.2, no.1, March 2012, pp. 60-62. AUTHOR Puneet Ranout is a Graduate in Information and Technology Engineering from Punjab Technical University, Jalandhar, Punjab in 2012. Presently he is pursuing Post Graduate (Final Year) in Computer Science Engineering from Uttarakhand Technical University, Dehradun. His area of interest include computer network and image processing. Anubhooti Papola is a graduate in Computer Science Engineering from H.N.B Garhwal University, Srinagar, Uttarakhand in 2009 and a Post Graduate in Computer Science Engineering from Graphic Era University, Dehradun in 2012. She was a Lecturer in GRD IMT Dehradun and programmer in Anya- Softek, Dehradun. Presently she is working as Assistant professor in W.I.T, Uttarakhand technical university, Dehradun. Devesh Mishra is a graduate in Computer Science Engineering from Agra University in 2003 and a Post Graduate in Information and Communication Technology from Dhirubhai Ambani Institute of Information and Communication Technology, Gandhinagar, Gujarat in 2009. He was Lecturer in Amity School of Engineering. He worked as Assistant Professor in Uttarakhand Technical University, Dehradun. Presently he is working as Lecturer in Department of technical Education, Rajasthan Government, Rajasthan Volume 3, Issue 9, September 2014 Page 314