A Color Image Segmentation Method Based on Improved K-Means Clustering Algorithm
How to effectively segment objects in the color images is the key point in the computer vision and image analysis. All kinds of segment algorithms have been proposed by many scholars, which could be basically divided into three categories. This paper pres
- PDF / 438,816 Bytes
- 7 Pages / 439.37 x 666.142 pts Page_size
- 75 Downloads / 225 Views
A Color Image Segmentation Method Based on Improved K-Means Clustering Algorithm Ran Jin, Chunhai Kou, Ruijuan Liu and Yefeng Li Abstract How to effectively segment objects in the color images is the key point in the computer vision and image analysis. All kinds of segment algorithms have been proposed by many scholars, which could be basically divided into three categories. This paper presents repeated usage of the optimal threshold for roughly extracting the largest target area of the color image. Then an improved K-means clustering algorithm is used to improve the accuracy of the segmentation from the target area. Experimental results show that this method can effectively extract color image from an object. It has also a certain degree of robustness to the noisy image. Keywords Image segmentation • Optimal threshold • K-mean clustering algorithm • Robustness
63.1 Introduction In the computer vision and image analysis, it has been the key point as how to effectively segment objects out of image. All kinds of segment algorithms have been proposed by many scholars, which could be basically divided into three categories. Category 1, named as threshold algorithm, which segments objects utilizing the whole or partial of the image. This knowledge is generally presented by histogram graphics. This tool is very popular and the easiest segment method characterized by fast speed processing and low cost. The threshold algorithm could automatically determine the R. Jin (*) Zhejiang Wanli University, Ningbo 315100, China e-mail: [email protected] R. Jin · C. Kou · R. Liu · Y. Li Donghua University, Shanghai 201620, China C. Kou e-mail: [email protected] R. Liu e-mail: [email protected] Y. Li e-mail: [email protected] Z. Zhong (ed.), Proceedings of the International Conference on Information Engineering and Applications (IEA) 2012, Lecture Notes in Electrical Engineering 217, DOI: 10.1007/978-1-4471-4850-0_63, © Springer-Verlag London 2013
499
500
R. Jin et al.
threshold value, which generally includes the P rate threshold algorithm, histogram contour analysis threshold algorithm and the optimal threshold algorithm, etc. [1]. The determination of threshold value and noise in the image signal will directly impact the efficiency of segmentation. Category 2, contour-based segment, which replies on the detector to locate the image peripheral and the located peripheral, presents the discontinuity of images in regard to gray value, color and stripe. Methods are commonly used to include threshold of image contour, blurry contour and Hough exchange segmentation, etc. Yet this segmentation may probably lead to errors like creating contours that do not exist and with missing image boundary that is supposed to be there. These errors are due to the noisy signal or other unnecessary signals in the useful image signals. Category 3, region-based segment, it satisfies the conditions of complete segmentation and conformity of maximum region. It features the ability of robustness to image noise signals. Methods are wi
Data Loading...
