Segmentation Techniques In Image Processing - IJSER
International Journal of Scientific & Engineering Research, Volume 7, Issue 12, December-2016ISSN 2229-5518304SEGMENTATION TECHNIQUES IN IMAGE PROCESSINGPreeti PanwarDepartment of Computer Science & Applications, Kurukshetra University, Kurukshetrapreetipanwar102@gmail.comGirdhar GopalDepartment of Computer Science & Applications, Kurukshetra University, Kurukshetragirdhar.gopal@kuk.ac.inRakesh KumarDepartment of Computer Science & Applications, Kurukshetra University, Kurukshetrarakeshkumar@kuk.ac.inABSTRACTImage processing is a form of signal processing. One of the mostly used operations of image processing isimage segmentation. Over the last few year image segmentation plays a vital role in image processing .The goal of image segmentation is to partition the pixels into silent image segments i.e., these segmentscorresponding to individual objects, natural parts of objects, or surface. The problems of digital imagesegmentation represent great challenges for computer vision. Segmentation techniques which are used inimage processing are edge based, region based, thresholding, clustering etc.In this paper, different imagesegmentation techniques have been discussed.Keywords: Image, Digital Image processing, Image segmentation, Thresholding.IJSER1. IntroductionImage processing is the general issue in today’sworld, in the field of computer vision. Imageprocessing is the form of signal processing whereboth the input and output signals are images. Animage may be defines as a two-dimensionalfunctional, f(x, y), where x and y are spatialcoordinates, and f is called the grey level orintensity of the image at that point. Images can betwo-dimensional signals via a matrix representation,and image processing can be understood by employone-dimensional signal processing techniques totwo-dimensional signals. Applications of imageprocessing are satellite imaging, medical imaging,photography, and image compression etc[1]Image processing basically includes the followingtwo steps: Importing the image via image acquisitiontools;Analysing and manipulating the image.1.1 Methods of image processing:There are two type of methods used for imageprocessing namely, analog and digital imageprocessing.Analog image processing refers to the modificationof image through electrical means. The mostcommon example is the television image. Analogimage processing is mainly used for the hard copieslike printouts and photographs.Digital image processing: - The term digital imageprocessing generally refers to processing of a twodimension image by a digital computer. Theprinciple advantages of digital image processingmethods are its repeatability, versatility, and thepreservation of original data precision. The threegeneral phases of digital image processing are preprocessing, enhancement, display informationextraction.2. Digital image processing:Digital image processing uses computer algorithmto perform image processing on images to improvethe quality of the image by removing noise andother unwanted pixels and also to obtain moreinformation on the image. There are fundamentalsteps in digital image processing. These steps areimage acquisition, image enhancement, imagerestoration, color image processing, wavelets icalprocessing,segmentation,representation and description, object recognition.[2] The fundamental steps of image processing areas follows:IJSER 2016http://www.ijser.org1.Image acquisition: Image acquisition isrefer as to acquire the image. Acquisition is
International Journal of Scientific & Engineering Research, Volume 7, Issue 12, December-2016ISSN 2229-5518as simple as an image that is already indigital form. Generally, image processingis involves by the image acquisition.2. Image enhancement: Image enhancement isamong the simplest and most appealingareas of digital. The enhancementtechniques are used to enhance the detailthat is simply to highlight certain featuresof an image.3. Image restoration: Image restoration isused to improving the appearance of animage. Image restoration techniques tend tobe based on mathematical or probabilisticmodels of image degradation.4. Color image processing: Color imageprocessing is an area that has been widelyused now days because of rapidly use ofdigital image over the internet.5. Wavelets: Wavelets are used forrepresentating images in various degrees ofresolution. Basically this is used forpyramidal representation and image datacompression in which images aresubdivided successively into smallerregions.6. Compression: Compression is a techniquethat is used for reducing the storagerequired for saving an image or thebandwidth required for transmitting it.7. ulinrepresentation and description of shape areextracted from morphological processingand it deals with image extraction tools8. Segmentation:Segmentation subdivides animage into its essential parts or objects.The level of subdivision is depends on theproblem being viewed.9. Representationanddescription:Representation and description follow theoutput of a segmentation stage, which usualis row pixel data, constituting either theboundary of a region or all the points in theregion itself.Description also called feature selection,deals with extracting the information thatresult in some qualitative information ofinterest. It differentiate the one class ofobjects from another.10. Recognition: - Recognition is the processof assigning a label (e.g. “vehicle”) to anobject based on its descriptors.[3]3053. Image segmentationAmong the various image processingtechniques, image segmentation is veryimportant step to analyse the given image andextract data from them[4]. Segmentation is aprocess to subdivide the image into small imageregion and that region corresponding toindividual surfaces, objects, or natural parts ofobjects. The level of subdivision is dependingon the problem being solved. That is,segmentation should stop when the objects ofinterest have been achieved.The goal of segmentation is to change andsimplify the representation of an image intosomething that is more useful and simple toanalyse. However, it is the process of assigninga label to every pixel in an image such thatpixels with the same label have somepredefined characteristics. All of the pixels in aregion are similar with respect to some propertysuch as colour, intensity, or texture. Someapplications of image segmentation are imageprocessing, medical imaging, computer vision,digital libraries, face recognition, image andvideo retrieval, satellite image. [5].Based ondifferent technologies, image segmentationapproaches are currently divided into followingcategories, based on two properties of image.IJSERIJSER 2016http://www.ijser.orgDetecting Discontinuities:-It divide an imagebased on short change in intensity, this includesimage segmentation algorithms like edgedetection.[1]Detection Similarities:-It divides an image intoregions that are similar according to apredefined criterion, this includes imagesegmentation algorithm like region [6]3.1 Edge-based techniques:
International Journal of Scientific & Engineering Research, Volume 7, Issue 12, December-2016ISSN 2229-55181.Edge detection is a major tool for imagesegmentation. Edge defines the boundaries betweenregions in an image. Edge detection of an imagesignificantly reduces the amount of data and filtersout unusable information, while keep the importantstructural properties in an image.It could detect the variation of grey levels, but it issensitive to noise. It is main tool in patternrecognition, image segmentation, and sceneanalysis.[7] Edge are local changes in the imageintensity edge typically occur on the boundarybetween two regions. The main features areextracted from the edges of an image. Edgedetection has major features for image analysis.These features are used by advanced computervision algorithm. Edge detection is used for objectdetection which gives various applications likebiometrics, medical image processing etc.[8]There are three different types of discontinuities inthe grey level like line, point and edges. Spatialmasks can be used to detect all the three types ofdiscontinuities in an image. All the edge detectionoperators are grouped under two groups are:1.2.2.3.4.306Select a group of seed pixels within animage.Select a set of similarity criterion such asgrey level intensity or color and setup astopping rule.Grow regions by attaching each seed thosehave predefined properties similar to seedpixels.Stop region growing when no more pixelsmet the criterion for inclusion in thatregion.[10]Region splitting and merging: - This methodsegment the image based on homogeneity criteria.This method works on the basis of quadtrees. Itconsiders the entire image as a single region andthen divides the image into four regions based oncertain predefined criteria. It checks the regions forthe same defined criteria and divides it further intofour regions if the test result is negative and theprocess continues till the criteria is defined. [4]IJSER1st order derivative: - 1st order derivativesare Sobel operator, Canny operator, Prewitoperator, Test operator.2nd order derivative: - 2nd order derivativesare Laplacian operator, Zero-crossing.[6]3.2Region based techniques:Region based methods are based continuity. Thistechnique divide the entire image into sub regionsdepending on some criterion like all the pixels inone region must have the same grey level. Thesimplest approach to segment image based on thesimilarities assumption is that every pixel iscompared with its neighbour for similarity check(for grey level, texture, color, shape).3.3Thresholding:Image segmentation by thresholding is a simple butpowerful approach for segmenting images. It isuseful in select foreground from background.Thresholding operation convert a multilevel imageinto a binary that is it chooses a proper thresholdingT, to divide image pixels into several regions andseparate objects from background. Any pixel (x, y)is belong to object if its intensity is greater than orequal to threshold value i.e., f(x, y) T, else pixelbelong to background. [11]Based on the selection of threshold value, there aretwo type of thresholding method:1.Region based technique is relatively simple andmore immune to noise as compare to edge detectionmethod. Edge based method divide an image basedon changes in intensity near edge whereas regionbased methods, divide an image into region that aresimilar according to set of predefined criteria.[9]Region growing: - Region growing is a techniquesfor extracting a region of image based on predefinedcriterion .Region growing can be prepared in foursteps:IJSER 2016http://www.ijser.org2.Global thresholding: - Global thresholdingis used when the intensity sharing betweenthe objects of foreground and backgroundare very distinct, a single value of thresholdcan simply be used to separate both objectsapart. In this type of thresholding, the valueof threshold T depends on the property ofthe pixel and the grey level value of theimage. Some of the common used globalthresholding methods are Otsu method,entropy based thresholding, etc.Local thresholding: -This method dividesan image into several sub regions and thenchooses different thresholds Ts for eachsubregion respectively. Some commonused local threshold techniques are 2-D
International Journal of Scientific & Engineering Research, Volume 7, Issue 12, December-2016ISSN 2229-5518entropy-based histogramstatisticalNeural networkimportant steps:Hard clustering: - Hard clustering assumes that apixel can only belong to a single cluster. Thereexistsharp boundaries between clusters. One of the mostpopular and well used clustering algorithms is kmeans clustering algorithm. K-means clusteringgroup n pixels of an image into k number of cluster,where k n and k is a positive integer. The centroidsof the predefined clusters are computed randomly.Clusters are formed on the basis of some similarityfeatures like distance of pixel intensities and graylevel intensity of pixels.includestwo(1) Feature extraction: - The input data ofneural network is determines in this step.Some important features from images areextracted.(2) Image segmentation:- The features that areextracted from the image are segmented inthis step.3.4Clustering:Clustering is an unsupervised learning task.Clustering is defined as process of grouping objectbased on attributes, so that object with similarattributes lies in same cluster. Clustering is used forthe purpose of pattern recognition, imageprocessing, data analysis, and more.Clusteringalgorithm is classified as hard clustering, k-means,fuzzy clustering, etc.[13]segmentation307Neural network have fast computing and highlyparallel computing ability makes it suitable for realtime application. It improves segmentation resultswhen the data deviated from a normal situation. It ishigh robustness that makes it immune to noise.[12]3. ConclusionDigital image processing uses computer algorithmto perform image processing on the image. Theoverview of various segmentation methodologiesapplied for digital image processing is explainedbriefly in this paper. Throughout this study ofvarious techniques, it is concluded that imagesegmentation is the crucial part of image processing.The segmentation technique of the image could beused as per the required application. Since imagesegmentation is affected by lots of factors such astype of image, color, intensity, level of noise, etc.thus there is no single algorithm that is applicable onall types of images and nature of problem.IJSERFuzzy clustering: - Fuzzy clustering can be usedwhen there are no defined boundaries betweenobjects in an image. Fuzzy clustering divides theinput pixels into cluster on the base of somesimilarities criterion. Similarity criterion can bedistance, connectivity, intensity etc. Fuzzyclustering algorithms include FCM (fuzzy c means)algorithm, Gaussian mixture decomposition, FCV(Fuzzy c varieties), Gusatafson-kessel etc.[14]3.5Artificial Neural Network:A neural net is an artificial representation of humanbrain. It tries to simulate its learning process. Inrecent years, artificial neural networks areused tosolve the problem of medical image segmentation.Neural network based on simulation of life,especially the human brains learning process. Eachnode can perform some basic computing. [15]In this, firstly the image is mapped into a neuralnetwork where every neuron is represented byapixel. The neural network is trained with trainingsample set. It helps todetermine the connection andweights between nodes. With the help of trainedneural network the new images are segmented.Reference[1] S. Saini and K. Arora, “A Study Analysis onthe Different Image SegmentationTechniques,” International Journal ofInformation & Computation Technology,vol. 4, pp. 1445-1452, November 2014.[2] J. Singla, “TECHNIQUE OF IMAGEREGISTRATION IN DIGITAL IMAGEPROCESSING - A REVIEW,” InternationalJournal of Information Technology andKnowledge Management, vol. 5, pp. 239243, December 2012.[3] R. Rafael C. Gonzalez, Digital ImageProcessing, 2002.IJSER 2016http://www.ijser.org
International Journal of Scientific & Engineering Research, Volume 7, Issue 12, December-2016ISSN 2229-5518[4] R. S. A. M. Khan, “Image segmentationmethods:A Comparative Study,”International Journal of Soft Computingand Engineering (IJSCE), vol. 3, no. 4,September 2013.[5] A. G. Amandeep Kamboj, “Simulink ModelBased Image Segmentation,” InternationalJournal of Advanced Research in ComputerScience and Software Engineering, vol. 2,no. 6, june 2012.[6] H. Narkhede, “Review of ImageSegmentation Techniques,” InternationalJournal of Science and ModernEngineering (IJISME), vol. 1, no. 8, july2013.[7] G. S. A. K. Pooja Sharma, “DifferentTechniques Of Edge Detection In DigitalImage Processing,” International Journalof Engineering Research and Applications(IJERA), vol. 3, no. 3, pp. 458-461, mayjune 2013.308[12] A. K. A. K. Rohan Kandwal, “Review:Existing Image Segmentation Techniques,”International Journal of AdvancedResearch in Computer Science andSoftware Engineering, vol. 4, no. 4, april2014.[13] G. D. Mohit Agarwal, “Application ofclustering technique for ImageSegmentation,” International Journal ofAdvanced Research in Computer Scienceand Software Engineering, vol. 3, no. 4,april 2013.[14] P. J. Shah Nilima, “Review on ImageSegmentation, Clustering and BoundaryEncoding,” International Journal ofInnovative Research in Science,, vol. 2, no.11, november 2013.IJSER[8] M. Muthukrishnan.R, “EDGE DETECTIONTECHNIQUES FOR IMAGESEGMENTATION,” International Journal ofComputer Science & InformationTechnology (IJCSIT), vol. 3, 2011.[9] P. G. Manjot Kaur, “A Review on RegionBased Segmentation,” InternationalJournal of Science and Research (IJSR),2013.[15] D. G. R. K. R. V. K. R. M. Jogendra Kumar,“REVIEW ON IMAGE SEGMENTATIONTECHNIQUES,” International Journal ofScientific Research Engineering &Technology (IJSRET), vol. 3, no. 6,september 2014.[16] H. H. Z. K. M. Cheng, “Final Project Report– Image Processing Techniques,” 2006.[17] J. Singla, “Techniques of ImageRegistration in Digital Image,”International Journal of InformationTechnology and Knowledge Management,vol. 5, pp. 239-243, december 2012.[10] P. S. D. Rajeshwar Dass, “ImageSegmentation Techniques,” vol. 3, no. 1,march 2012.[11] N. K. K. S. Salem Saleh Al-amri, “ImageSegmentation by Using ThershodTechniques,” JOURNAL OF COMPUTING,vol. 2, no. 5, may 2010.IJSER 2016http://www.ijser.org
Department of Computer Science & Applications, Kurukshetra University, Kurukshetra . rakeshkumar@kuk.ac.in . ABSTRACT . Image processing is a formof signal processing . One of the mostly used operations of image processing is image segmentation. Over the last few year image segmentation plays vital role in image pra ocessing .
The input for image processing is an image, such as a photograph or frame of video. The output can be an image or a set of characteristics or parameters related to the image. Most of the image processing techniques treat the image as a two-dimensional signal and applies the standard signal processing techniques to it. Image processing usually .
Image segmentation is one of the many image processing algorithms. It is used mainly to reduce the original image data content for further processing. Image segmentation basically partitions the input image domain into regions, and each region contains pixels with a certain similar property with respect to each other within the region.
Methods of image segmentation become more and more important in the field of remote sensing image analysis - in particular due to . The most important factor for using segmentation techniques is segmentation quality. Thus, a method for evaluating segmentation quality is presented and used to compare results of presently available .
Internal Segmentation Firewall Segmentation is not new, but effective segmentation has not been practical. In the past, performance, price, and effort were all gating factors for implementing a good segmentation strategy. But this has not changed the desire for deeper and more prolific segmentation in the enterprise.
Internal Segmentation Firewall Segmentation is not new, but effective segmentation has not been practical. In the past, performance, price, and effort were all gating factors for implementing a good segmentation strategy. But this has not changed the desire for deeper and more prolific segmentation in the enterprise.
segmentation research. 2. Method The method of segmentation refers to when the segments are defined. There are two methods of segmentation. They are a priori and post hoc. Segmentation requires that respondents be grouped based on some set of variables that are identified before data collection. In a priori segmentation, not only are the
Fig. 1.Overview. First stage: Coarse segmentation with multi-organ segmentation withweighted-FCN, where we obtain the segmentation results and probability map for eachorgan. Second stage: Fine-scaled binary segmentation per organ. The input consists of cropped volume and a probability map from coarse segmentation.
Web Hooks and API integration Add validations and extensions from the marketplace or build your own using web hooks and REST APIs. Semantic code search Quickly find what you’re looking for with code-aware search that understands classes and variables. Getting Started with Azure DevOps Azure Pipelines Cloud-hosted pipelines for Windows, macOS, with unlimited minutes for open source page 013 .
