A Survey on Image Segmentation and its Applications in Image Processing

Page 1

Int. Journal of Electrical & Electronics Engg.

Vol. 2, Spl. Issue 1 (2015)

e-ISSN: 1694-2310 | p-ISSN: 1694-2426

A Survey on Image Segmentation and its Applications in Image Processing Kamaljeet Kainth1, Amritpal Singh2, Priya Chitkara3 1,2,3 1

Dept. of ECE, Guru Nanak Dev University Regional Campus, Jalandhar, India

kainth108@gmail.com, 2asdhunna91@gmail.com, 3chitkara23@gmail.com

Abstract: As technology grows day by day computer vision becomes a vital field of understanding the behavior of an image. Image segmentation is a sub field of computer vision that deals with the partition of objects into number of segments. Image segmentation found a huge application in pattern reorganization, texture analysis as well as in medial image processing. This paper focus on distinct sort of image segmentation techniques that are utilized in computer vision. Thus a survey has been created for various image segmentation techniques that describe the importance of the same. Comparison and conclusion has been created within the finish of this paper. Keywords: Image segmentation, Image processing,Fuzzy-C Means (FCM) ,Edge based segmentation.

I. INTRODUCTION Image segmentation plays an important role in computer vision. Pattern recognition, texture analysis, medical image processing, facial detection and many security system deployed segmentation technique. These segmentation techniques are broadly classified into following categories, Edge based segmentation, Fuzzy based detection, PDE based segmentation, Region based segmentationThreshold based segmentation .Edge detection is one of the major tool for image segmentation. Edge detection is the approach for detecting significant discontinuities in intensity values. In gray level main discontinuities are point, lines and edges. In order to find the discontinuities these techniques utilize are Robert edge detection, Log Gabor edge detection [1], Sobel edge detection, Prewitt edge detection and canny edge detection. An image is segmented on the basis of gray level discontinuities which are present on the boundary the image [2].Fuzzy based detection applies the principle of clustering. Clustering is the process in which similar objects are grouped into one cluster. Fuzzy clustering is one step ahead to hard (crisp) clustering and mainly employed for acquiring fuzzy patterns. Fuzzy clustering is further categorized into fuzzyc means (FCM) and fuzzy kernel c-means algorithms (KFCM) [3]. In FCM distance between data points are considered to form a cluster and on the basis of this cluster centers are created [4]. However KFCM overcomes the disadvantage of FCM and kernel information is added to FCM, to cover small differences between clusters. Partial differential equation (PDE) based segmentation is an efficient technique for image segmentation. PDE based model is a geometrical active counter model which utilize fuzzy classification and can handle variation in topology of shape [5]. This model overcomes the disadvantages of fuzzy based detection hence a better option for image segmentation. Region based segmentation employ the approach of dividing the image into small regions. These regions are created on the basis of color, texture, intensity 83

value. The segmented regions are assigned with a label. Region based segmentation uses region transformation or histogram method. This approach provides advantage of continuous segmentation maps of closely related objects [6]. Threshold values for detecting corresponding area are enforced for efficient segmentation.Threshold based segmentation method is the simplest method for segmentation. In this method threshold values are obtained by histogram of the edges of the image. Global thresholding, variable thresholding and multiple thresholding technique are employed to select the threshold value. A threshold selection component is defined to select the threshold value which is automatically updated according to the contrast of the edges detected. But this technique has a disadvantage that this cannot be applied to complex images as partitioning of pixels is difficult [7].This paper will survey all the techniques employed for segmentation purposes. II. LITERATURE SURVEY Shinn-Ying Ho and Kual-Zheng Lee[8]. , proposed a novel approach which satisfies the basic objectives of image segmentation i.e. an algorithm must have an efficient segmented contour , computation time must be very small without setting any threshold and segmented regions should be segmented robustly. This efficient technique is termed as evolutionary image segmentation algorithm. This technique employ K-means algorithm which is utilized to split an image into several homogeneous regions. This particular algorithm also overcomes the issues which are faced by traditional image segmentation techniques like, over segmentation, continuous contour. The overall algorithm is split into two major categories named as split procedure and merge procedure respectively. First one groups the pixels of small region into the larger adjacent region. Second one utilized binary chromo sense channel encoding and further dived into seven steps to achieve the above mention objectives. The whole scenario reduces the computation time, improves the quality of splinted image and hence results into efficient robust segmentation techniques in comparison to traditional segmented technique.Various threshloding techniques have been developed for image segmentation in [9]. H.D. Cheng et.al in [10], presented an image segmentation technique which is based on homogram thresholding in conjunction with region merging technique. First one consider local and global information whereas later one finds peaks of homogram .Region merging gives free space and spatial relation between pixels. Further this paper compare histogram thresholding approach their approach which proves that afterward one estimate more information about gray level NITTTR, Chandigarh

EDIT-2015


Turn static files into dynamic content formats.

Create a flipbook
Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.
A Survey on Image Segmentation and its Applications in Image Processing by IJEEE (Elixir Publications) - Issuu