Review of Image Enhancement in Spatial Domain

Transcription

1 Internatinal Jurnal f Cmputer Science and Infrmatin Technlgy Research ISSN X (nline) Vl. 2, Issue 4, pp: ( ), Mnth: Octber - December 2014, Available at: Review f Image Enhancement in Spatial Dmain Vimal Bibhu 1, Bhanu P. Lhani 2, Megha Vaid 3, Anjana Swanny 4, Kmal Verma 5 Amity University, Greater Nida, Uttar Pradesh, India Abstract: Digital Image Prcessing is a prcess in which digital images are prcessed by the means f digital cmputer. Its main applicatin is t imprve the pictrial infrmatin fr human interpretatin. Almst all f the digital images cntain nise. T imprve a digital image, a number f enhancement techniques are available. Filtering is ne f the cmmn techniques used t remve unwanted infrmatin (nise) frm the image. Image enhancement plays a vital rle in every field particularly where images need t be clsely examined and understd. Different images like satellite images, medical images, aerial images, micrscpy images and even the real life phtgraphs cntain pr nise and a difference in lightness, brightness and/r hue between tw clrs that makes them mre r less distinguishable. It is als used fr image sharpening and smthening. The prblem f is cnsidered as a quality imprvement prblem. Digital images cntain varius types f nises which reduce the quality f images. Nises can be remved by varius enhancement techniques. Digital images can be either spatial dmain r frequency dmain. This paper investigates varius techniques used in spatial dmain image prcessing. Keywrds: spatial dmain; filter; smthing; sharpening; nise. I. INTRODUCTION An image is an array, r a matrix, f square pixels (picture elements) arranged in clumns and rws. A Digital image is a 2D light intensity functin in which each element is called pixel. Fig 1: An 8 bit Digital Image When an image is acquired by any imaging system such as camera, the visual system fr which it is planned is unable t directly use it. The steps invlved in Digital Image prcessing are image acquisitin, image enhancement, image analysis, image recnstructin, image restratin, image cmpressin, image segmentatin, image recgnitin, and visualizatin. In digital image prcessing, the majr surces f nise are at the image acquisitin and image transmissin. Research Publish Jurnals Page 194

2 Internatinal Jurnal f Cmputer Science and Infrmatin Technlgy Research ISSN X (nline) Vl. 2, Issue 4, pp: ( ), Mnth: Octber - December 2014, Available at: Image prcessing is a prcess which transfrms a degraded image int an image f better quality s as t facilitate its later interpretatin. Image prcessing techniques can be then applied in rder t facilitate the interpretatin f images by further using certain cmputer visin algrithms. The majr applicatins f image prcessing are fllwing: Phtgraphy and printing Machine Visin Satellite image prcessing Micrscpe image prcessing Medical image prcessing Face detectin, feature detectin, identificatin Car barrier detectin II. METHODOLOGY Image Enhancement is the imprvement f digital image quality (wanted e.g. fr visual inspectin r fr machine analysis), withut the knwledge f surce f degradatin. If the surce f degradatin is knwn, ne calls the image restratin step. Bth the icnic prcesses, viz. input and utputs are images. The enhancement appraches fall int tw brad categries: 1. Spatial dmain methd 2. Frequency dmain methd Spatial dmain methds directly mdify the image pixels t achieve desired enhancement in spatial dmain. Frequency dmain methds perfrm the enhancement peratins t Discrete Furier Transfrm (D F T) f an image in frequency dmain. Inverse Furier transfrm f spectrum is fund t get the enhanced image in which pixel values will be mdified. Image enhancement techniques in Spatial Dmain Sme f the cmmn image enhancement techniques in spatial dmains fr grayscale images are: 1. Pint prcessing peratins 2. Spatial filter peratins 3. Histgram prcessing peratins 4. Pseud clring peratins I. Pint Prcessing Technique In this technique, we wrk n asinge pixel at a time i.e. T is 1 x 1 peratr. The new image frmed after depends n this transfrm T n the riginal image. Research Publish Jurnals Page 195

3 Internatinal Jurnal f Cmputer Science and Infrmatin Technlgy Research ISSN X (nline) Vl. 2, Issue 4, pp: ( ), Mnth: Octber - December 2014, Available at: Sme examples r techniques f pint prcessing are as fllws: Digital Negative Cntrast Stretching Threshlding Gray level slicing Bit plane slicing Dynamic range cmpressin 1) Digital Image Negative: It is useful in a number f applicatins. eg. X-ray images. The term Negative means inverting the gray levels. Digital Negative can be btained by: s = 255 r (where, r max = 255) when, r = 0; s = 255 & if r = 255; s = 0 Generally, s = (L-1) r where, L ttal number f gray levels (e.g. 256 fr 8-bit image) 2) Cntrast Stretching This is dne when the captured image has pr Illuminatin r lens aperture has been placed wrng. Idea behind cntrast stretching is t make dark prtin darker and bright prtin brighter. Frmula is: s = l.r ; fr 0 r a = m(r-a) + v ; fr a r b = n(r-b) + w ; fr b r L-1 Research Publish Jurnals Page 196

4 Internatinal Jurnal f Cmputer Science and Infrmatin Technlgy Research ISSN X (nline) Vl. 2, Issue 4, pp: ( ), Mnth: Octber - December 2014, Available at: 3) Threshlding Extreme Cntrast Stretching yields Threshlding. If r1 = r2, s1 = 0 & s2 = L-1 Then we get threshlding functin. Expressin becmes: s = 0; if r a s = L 1; if r >a where, L is number f Grey levels. Threshlded images have maximum cntrast as there are nly BLACK & WHITE grey values. 4) Grey Level Slicing Threshlding splits the image in 2 parts. But at times we need t highlight a specific range f grey levels.eg. X-ray scan, CT scan. Grey level slicing lks similar t threshlding except that we select a band f grey levels in this technique. 5) Bit Plane Slicing In bit plane slicing, each bit cntributes t the frmatin f final image. Let us cnsider a 256x256 image with 256 grey levels. Fr eg- Black is represented as and White as Steps t find grey level: Cnsider LSB value f each pixel & plt image. Cntinue till MSB is reached. All 8 images will be binary. Observing the images we cnclude that Higher rder images cntain visually sufficient data. Lwer rder bits cntain suitable details f image. Example f bit plane slicing is stegangraphy 6) Dynamic Range Cmpressin (Lg transfrmatin) Smetimes, dynamic range f the image exceeds the capability f display device. Sme pixel values are s large that the ther lw value pixel gets bscured.e.g. stars in day time are nt visible thugh present due t large intensity f sun. Thus dynamic range needs t be cmpressed.lg peratr is an excellent cmpressin functin.thus, Dynamic range cmpressin is achieved using lg peratr. Frmula used : s = C.lg(1 + r ) where, C nrmalizatin cnstant & r input intensity Research Publish Jurnals Page 197

5 Internatinal Jurnal f Cmputer Science and Infrmatin Technlgy Research ISSN X (nline) Vl. 2, Issue 4, pp: ( ), Mnth: Octber - December 2014, Available at: 7) Pwer Law Transfrm Pwer Law Transfrm can be used t increase dynamic range f image. II. Spatial Filtering Technique Spatial filtering term is the filtering peratin perfrmed directly n the image pixels. Theprcess simply cnsists f mving the filter mask frm ne pint t anther in an image. At every pint (x,y), therespnse f the filter can be calculated using certain predefined relatins. Spatial filter can be classified int i) smthing spatial filters and ii) sharpening spatial filters. These filterscan be either linear r nnlinear. In linear filter each value f pixel in the utput image is a weighted sum f thepixel in the neighbrhd f the crrespnding pixel in the input image. Nnlinear filtering peratin is based n the pixel values in the neighbrhd. Nnlinear filter can be effectively used in nise reductin Smthing Spatial Filters Smthing filters are used fr reductin f nise and blurring peratins. Each pixel f a digital image is represented by red, green, and blue chrmatic intensities. Blurring can be used as an initial step fr ther image prcessing peratins. By taking neighbring pixels int accunt, extreme nisy pixels are filtered ut. Unfrtunately, extreme pixels can als represent riginal details, which were lst due t smthing prcess. Sharpening Spatial Filters The main aim in sharpening an image is t highlight every fine detail f the image, r t enhance the blurred pixels (perhaps due t nise r mtin). It is dne by adding the edges t the input digital images. Research Publish Jurnals Page 198

6 Internatinal Jurnal f Cmputer Science and Infrmatin Technlgy Research ISSN X (nline) Vl. 2, Issue 4, pp: ( ), Mnth: Octber - December 2014, Available at: III. Histgram Technique It is a graph f the frequency f ccurrence f each gray level in an image. If nk is the number f pixels having a gray level rk, in an image f size MxN, then prbability f ccurrence f rk is given by Where k=0, 1, 2 L-1 and fr a nrmalized histgram k varies frm 0 t 1. Even thugh histgram f an image cntains n spatial infrmatin, image prcessing peratins can be dne based n histgrams. Fig. Original image and its histgram equalized image alng with histgrams IV. Pseud Clring Technique Humans are able t distinguish many clrs ther than grey levels. Greyscale images can be artificially clred by using a prper clr map. Pseud clring prcess assigns clrs t gray level values f pixel based n specified criterin. This clring prcess highlights the rainfall levels as shwn in Fig. Research Publish Jurnals Page 199

7 Internatinal Jurnal f Cmputer Science and Infrmatin Technlgy Research ISSN X (nline) Vl. 2, Issue 4, pp: ( ), Mnth: Octber - December 2014, Available at: III. CONCLUSION In the time f digital images, the majr issue is the huge amunt f data. Statistical parameters can prvide infrmatin abut an image using just a few numbers. Many statistics are used t characterize image qualities like brightness, cntrast etc and are used as a basis fr digital image prcessing algrithms t imprve the quality f displayed digital image. Lcal enhancement methds discussed in this paper using statistical parameters are used t enhance images in every area. Mean and variance methds are used as the basis fr making changes in each pixel in the image. This helps in enhancing details ver small areas in an image. The images enhanced thrugh this methd are mst satisfactry when bringing ut the desired detail is cncerned. Future Scpe There is ample scpe fr future wrk in this research wrk t further imprve the perfrmance f the Lcal Image Enhancement Methd. Autmatin f this methd will help getting the cmbinatin f satisfactry parameters faster. Adaptive apprach t neighbrhd calculatin will increase the efficiency f image enhancement prcess. REFERENCES [1] E. Umbaugh, Cmputer Visin and Image Prcessing. A Practical Apprach Using CVIP Tls, Prentice Hall, 1998 [2] James Church, Dr. Yixin Chen, and Dr. Stephen Rice. A Spatial Median Filter fr Nise Remval in Digital Images [3] AnkitaDwivedi, Archana Singh, An Advanced Filter fr Image Enhancement and Restratin, OJAET, Vl.1, N. 1, 7-10, May, [4] Changhng Wang Tayi Chen ZhenshenQu, A Nvel Imprved MedimFilter fr Salt-and-Pepper Nise frm Highly Crrupted Images, Internatinal Sympsium n Systems and Cntrl in Aernautics and Astrnautics (ISSCAA), Harbin, China, 10 June 2010, [5] S.Sridhar, Digital Image Prcessing, Oxfrd University Press, [6] Stella, Dr. Bhushan Trivedi, Implementatin f Order Statistic Filters n Digital Image and OCT Image: A Cmparative Study, Internatinal Jurnal f Mdern Engineering Research (IJMER), Vl.2, Issue.5, and Sep.-Oct [7] Dr. M.Renuka Devi, S.Shenbagavadivu, A COMPARATIVE ANALYSIS OF MEDIAN FILTER ON VARIOUS TYPES OF IMAGES, Internatinal Jurnal f Cmputer Applicatins(IJCA) ( ). [8] S. Sangwin, R. Hrne, The Clr Image Prcessing Handbk, Chapman & Hall, 1998 [9] R. Oten, R. de Figueired, Adaptive Alpha Trimmed Mean Filters under Deviatins frm Assumed Nise Mdel, IEEE Trans. Image Prcessing, Vl. 13, N. 5, 2004, pp [10] Research Publish Jurnals Page 200