@SdobLihid voit-+ Girenge Bepmentalion '- Imag. aryrudabion Aubabnais : an ammage nin ols Lon stitunt sugions ‘on’ objath- for egret dwel ptolwur eatradion , mage An qmantahion € We Junaamental For image ugmantalion we depend upon qealus oy an tmage auch a point Sui aye and -—. Image axgmentation va vouE ampudont but AY rate puaress becouse the accursus of vinage 41gmanlaition the auccens oF fault of Lomputuired peedunt of an wage. stion!- “foint olshution i baud on disSo these. duololy polnta ,whodt anhinsihy ou Wr ruplly roy ba dutertid 4 traight yorwanh by using Reali that tue general expsussion fan suspense of Oo filth /mark is given by R wiz, +W92.2.+ ~~ 24 R= e wrzt toh, Zr a _e duel of tre prael Quociahd Wi. Th. susponse of theme @ ith Center decahdnLine Detestion- 4 ud alin baud on abaupt ane tn ntunsihy of quay duel of umage Ta ola yl OQ discontinuity uh amage.” ! for See tine duituation we Ud tht Yollowing mask . | Be El Ge ey yy Ee eS a | ic Horizontal Vodscal Pg By moving Yost mark on the dimagt wwe con dilit the one -pixel thick uns ath hosizontal clinethidn - she dimection of Linu 4 abenpliy defoled lr bhe W-ejficients °F mask: Now duppoat fhok we have sunall th maiks on sieage © and tha au four sursponses RirRe aks and Kr upper shat at o partialar point IR] whew j= sat andj]Re. Edge _deluclion |- Edge dufuchdn va ont of dhe most — um posstant phenomena. an amnage paroceasing . Tntroducliin and Basics of Edge dilubiin'- € dge dabuhon Js alu baatd on dicontinuily meauursment. r+ alan shows she abruph Change Jn the Jntensily of quay duvet Esge tan Re dufined as 4 Lut of Lonnettid pixel that gy dies sin dhe boundawy butween two suaglong "An idol edge is defined as a att of Lonnethd pixels each of whith docatid at on Oxthogonal typ transition Jn quay luel” Te key ponds one;- » 4. The Slope of samp ws dnveutly Proportional ut clegree of blunsing un the edge . Mos blunting Means fui Lope. z .3. Bluwud edges KL oLuvasys hick and thasp talges aneEdge Deudori- Ad Jne beginning of edge shal eclges tn the guoy fuses at the DO of two xregtors uo We can ux chrivatives yor ht meawsment of ecg Ae dy drow 4 no edge meant no ity duel , pu suilant cleatvadive of image au. absupt change Change sh untens' of dhol coun will be zed Tf denvelive thows some dru 0 possibiliby of edgu- @ value means, ate of practical edge The fsut and aseond olexiv Bs we Row that hamb J given by gi - nt) =t Ao faut durivabie wilt & digit) _ orlt) 4 e pe dt edge always Wows ao Conslant just olnaivals el2. Now dake the treond olenivalrve Hoe Wwe dee , we yt positive tuond durivedin ox edge at didi. ancl rugative Jor edge. ot white dike, Tha by Auond aarivatile we can Jcnd the edger 4itualed a daxk, arde and white tide of umage a /| a aeFinal Dosivative COxadiont Opewatior ) \- fisat duvahie wi alao caltid quadient of an signal mathematically, . The gradient o} an timage dng) of docation (x43) ib difthed a she vector ot My -[ | 1 pan by] =} og 4 We Rnow the gradient Lows ght maximum jab of Change at any point ) Magnidude of this vecton is defined as meg (9f) = [Gs +6) * This quantiby gives the max sath of Change of tly) io ft unit distance sh che ductor of V4 Let olay) supruud the ang. of veeton Vf arta then nis ong given by i 5) (ox j (55) baits of JowtSobel @ Robe (ross gradient opexatiny j— : (ay = (Za-2s) Gay = (Ze-Ze) » frawitt Operaton Pruisl operator is perfarmad Suing mask , the durivabive uh y diceetion fist + tevid YowGix= (21420424) - C2) tZ4 424) — (Sobel opexator !- Sobel opexaton is a abightly modified veriton of pauwitl operaton and gcven by Gx = (21422429) -(2142 Zat2s) Gy = (24 4920424) ~ (Zit 224 427) e Second Nvuvatuie - Second oluu'vative vi olan Callid Laplacian ond dyed by Nol = be * oy Atwond dyuvaty Js ato Cold Japtacion and defined St p34 4, 3 eo. fon tsodropic rohahion Uinuu ment of #5° uk ik given by va = A25-(2a+%+ +Ze+Z0) :But Laplacian Jundudo ot not wud dn je sual 407 becawit A has tha following disadvantages” (@ Moxe Sensitive to Noise fe Td will provide cloubli edge . do Lontuaing dome Lime for edge duteation eo fe) Td will not give thr disucbuin of edger Jo we uit Joplacian Bion! with othu Amoi function go susluand the noite Generally daplasian sa utd with gaussian funclion dhat vis qélen by. e =a his Waray and « = Deviatwn SD ™Ala)=“Edge dtinhing’- Edge aston yield pines ah Bnage ie on edger OW nerd atep us to Say wee Hust pixels dngdher vnto a aut of edges Tn otha, words we may doy that our aim Ja to Haphace Pointe on edgu with a Jew edges Rastally “the need of edge dinbing us due to the | dollowing dw? suacons’, A. Small pieces of edgu may be missing S. Smoll edge aUugminls mai, appear vs ti pruaint due Jo noise Whee shou Li no sual edge Tn genoa edge Jinklng methods Can be Classified dato | two catiguuits (0) Local Edge Linkewsi- (adobat Edge Linkow'- ink |- On of tht dimplust abbroash dor= Now we have to odecuss the cuideuia of similanition in of analyai. Tha baal culexia b th Jype upon. Of the gradient opexator 4. The dbungth of the Wud do produce the edge fixel The olixection of qxadiend vero fox a patirular point Lxorys) Ty Alsungth of gradient = meq [ Cot) Ctortey] 7 Divtectidn of guadient = w(Xos4.) for 0 point (Hoy) at dhe edge Staingth of qxadrent = meq [yexyy] Disuchibn of gradient = « (try) Point (xe,ys) B told Jo similan oxy) 4 e meg Cf Coa) mgt tltaydl]gE E ! Dk Bena Pink atl 2 imines (oriole sere yt Oh aa on pindHoh Transform > Tt kuad b Lonnect dy) eit edge poids ® Cquakdn of dun Ys mac whit m- Uopt C = intercept } tee ae =e A dings poi tan be pod Thuufore Wt drennsjocm thot poict: * Sai pow ais a dou in ju ot (ccs) > Comat “b) > Conde) aia % ==> Math ematical manpnote —Juinear processes wonich h to sumove dete wa a collection of non ‘edrespi op exten ou sina & eae a Abuct elemurd ove aa ee modafced win Lush 7 shat 2 centned vex on fie plac ot Lome point. Whin Sha Atulrg dument OO (enbud over & i of tat sagt a sogteah oprxalion ba Ke on the pirds toveud ee 7 Basie eee 8 tee _opuraiiion | - 1. Union Ave = dy {xixenorxe®} Y > a 1- ANB = ANS “ is 3. Diffuunsed AB 5 { z)s8 andxceo) r]x eAand xéB]Compliment i AS did f zx4a} ae ae " 7 (| , “ RefLectian; > 8 = { wlw--b dr bes} “ah i> 2G | e Touan slationt 2 e A= {cle=arz for rer} Tis te opescaltdon ;- oR eee: > onion oF + fies) > * pO is Atl do drain bie [e] os 5 ||: Meripho toatl. operator! § Dilation |» Dilation i a paoceas wn wahich tint bi may is @xpardid from ‘fs ongcal thape, Tra way de o vimage. ss Cxparrdlid ia Adlumingd, sy dha Quuctwiing clement The abructusl darunt A amallen By ace vimposud do thi simaaye Lat are ape @ He dice Watd por Aaueliwny dered wa e er Atk A and @ tn 2 unas ‘ (means an umage ) He etosion of A BS 9 denoLiol by AOB L dufuted a } Aes ={ Z|(8)2cA3eqn ahows that “the Uwosion of A dy B ot Of alt points uh z auch dal 6, Anifled by 2 Contained un A ! Jl My hes. e > Abblitcalian of erosion tb Jor ehirminaling isnelevent size aatail Jxom Bay bina Image tx Human booty as Skeleton. BA = fF Uh9) Un) -0) C009 C9) C89) 8) Crag Be § (00) »0,0)3 e 2 i Lisp) C1121) 29 C00) (1,3) 6213) (3, 39099 ,Ly0) (an C2) Cs) C2) O49 043) Cota?eument u iwmpoud wojth he dv Gut aunpuk a da ae = dumerd ix overlapped with dhe lapel smagl ng f eager a ; p dhe Jap pirel. duiud oN: * oe duintd OFF. thug aur 70 pixel Ove Aolutton- Te duunduaing "an " TY all te pl loping belvern J ye sipuk vinage do the oullput whown aulow for Joust ibe addr In fied dow , the att image mot poyedttis ovedapped i with ye na alk de pl u) Oh OFF. output Snagh a ant dict 20Sepa Now dhe i clement 1m ve angprd vow of Uae input airages true ib ne poet en “ct i on: pl Lowerd the juight aye one dae So Wards Wk ts » we yond et maloh » and hanes dist ofp amaigt Is Auinto black bore With thouon below — Ath Ke above pwiers Jupeated “DU te poel eee yt 7s taben get acopurd - i eyed punndsi- BiTason and wrosion ( pens? - Morphategcal opening of an Smage. a bavicaliy 7 q asiogsqollowrd by dikapion using dhe dame ety cLument Te A tem Srage and bh 4 te Muuctving Clara han” t ow y ery oh Amys guen %. ; open (Avs) = DC ELA) 3D — Dileboo & + Crosion. aio Written a+ Opening —n Amooths du. Contours ef the airage DP bousks down naw berlolyes and eb'minals thin ¥ pote sions Tse a. 3 opabor one nage below & mage Jp ‘of ice loxte in Ht _or oc Closing :- monphole gal Uubuiry of an sina * aidabon placed by ensaion wary ubutluw'g — urent Cros, CAO) = ELDLA)) alan woillin a4 ’ Arg ~(AOBOS ‘ é Closing yenerally. a pari dlomnt mall holu. Tu aim purpose of assesing, Ste dpeantiim 05 partial . e puapcess Q Puyon She operthg opouafion one amass simon wae ating erect EE ay 9) te a2 a omy 7 Open (A.8) = DLE (a) pew = (R0S)64K E (A) (vusion)Aes - faefojepsieyyesete) eltib el iu Gh yeleiife output snage . (as), » : e Guu: perform Gosieg opvuation on ne Lmnige wthown bul, The size of He roger 4 Joxio . The tooo Loki block @u not cnnethd of Ye consider 'r Lon nectiny Hatha& C08) Ww sion abb.2 ° 3 ° ° ° ° ° ° ° ° ° 2 ° ° °ke ye izie jae ey GAYHit om -Miss Transorm alton | ~ Li iia ea a ool a3 ae j. 2 ; (PSalw-x) oT co BM, \ (A©x)A (AS OCW») joint xumop Cut) JnA@B= (Aor)n[ Ace (w-n7 * Generalized notation: § . (6,81) ’ Bit at Jormeol 4uom elimentr of 6 aaa ocialrd with an object. 7 Ba: tut Yormid Jaom umn of © oasoutated eatth the CasALa ponding bask ground. Bae X anol Bs = (W-xJ © Mow gqenenal olufenition;- (A*8) = @ea)nlA&e Br) A@B Conlon all the orihin point at which y Simullane By Yound a hit inf ard 1 Jound nu to AS e © Simple example |— Ly SsoLated Prpround pieels +No mnusiphbourtng fuuground pixel, L fasuugaourdl endl points & ound end points points.Key wods- olilation Monphotngical alywetthms— tonen dealing with bingy winages , the poulncible. appiittafion of morphology, a extrading nage omporunts hot au uutful ah the Jupusintation and adiscriptian of what Seema eetstxction!- The poundany extxacbon BLA) of a aut A tb a. | pla) =f -(AeB) Whos 8 a a wild oie clumant yy B=: ada] shen flow would br dame ay 1 i e aA. a except on pinch bn quom abl the stds. Hens A- ROA) would qe wa aris one pixel dianeA- hOB Henn.yg Siig: 2 ga a) < tne age Ade tacplimaiitain wd wnluabin: 0 Se a” = (Xr, 68) 0 AE Rae Hew %o = J i! poe Bs abuuduuiy elerard - alganten inaty whan Xk = Ket "7 igen Ba oy et osayian ply te FE ° du Sag, Tou aang srr ui alan glen. ou cnsidund ar 4. nae me tedBep=t f) ‘| qZ H pp J | es | Hi : TEE 1 © Ges —_% = Ose’ Birney xs,ne ors Coluulatid by abre poocets @ eee atop XK eky Now Final Jap - se| > i evga 4 Sinodin of oad Lomborend . Xee Xv 68) AA Kelso Convex Hull! Lat a ve ae oval Seay ond LA BF a Jour abated lament dhowon faclour & By bs > Uh Pu SE molt wx x xa ee eh)ue - Peta Hon x4 ot 2x! Then Convex hut or pis Ao crap = UD ‘ vs nig AOR = An Cant sl tsie tel23+, Galea meWaberahud_Legranlalure = Wattesad A adhe Kalntoahar o7 — al qe aurmedtil on tr arta of ewuh- ox Jand vwuttesh fon Qe a lea rage un be agrntd se sugion (ohare tarald low drt du Jarne Laloc eglonal minimum,» Ridge wien) pinta shud Lint . : s point ah whith a chop of wialer wilh be equal Lilly Jo oul da mow. dhanone minimum. Goala_o} the algorithm |- O find dhe walerebel Lr 4 ; Aagreslation by watuukeol embodlis many et Concept -0} the othr three a} prema ard , os — hah ‘often produces move ttoble Arg mentation ig tonnettio Suymentaten bounda a dimbie 4on these conce The principal objective of segr is to find the watershed lines. asic ideu is simple, a illustrates. Suppose that a hole is punched in each regional minimum and th the entire topography is Nooded fron) below by letting rise through holes at a uniform rate, When the rising water in distinct catchment i ent the merging, The flooding about to m tually reach when only the tops of the dams are visible abo’ " line. These dam boundaries correspond to the divide lines of the w: There: ey are the (connected) boundaries extracted by a wa mentation hm, be ed further with the aid of Fig. 10.54. Figu i ew. in which age and Fig. 10.54(b) is a topogr ains” is. proportional to intensity in the rpretation, the backsides of structures are shaded. Th topography o to prevent the ris . we imagdams of height is determined by the perimeter of the entire topography (image) being en greater than the highest possible mountain, whose valu highest possible intensity value in the input image Suppose that a hole is punched in each regional minimum [sho areas in Fig. 10.54(b)] and that the entire topography is flooded from letting water rise through the holes at a uniform rate. Figure 10.54(c) shows the first stage of flooding, where the “water,” shown in light gray, has co areas that correspond to thé very dark background in the image. In Figs. 10 and (e) we see that the water now has risen into the first and second c: basins, respectively. As the water continues to risc. it will eventua’ Jy overflow from one catchment basin into another. The first indication of this is shown im 10.54(f). Here, water from the left basin actually overflowed into the basin on the right and a short “dam” (consisting of single pixels) was built to prevent water from merging at that level of flooding (the details of dam building are dis: cussed in the following section). The effect is more pronounced as water continues to rise, as shown in Fig. 10.54(g). This figure shows a longer dam between the we catchment basins and another dam in the top part of the right basin. The la dam was built to prevent merging of water from that basin with water from areas corresponding to the background. This process is continued until the maximumae oe ee evel of flooding (corresponding to the highest intensity value in the image) is reached. The final dams correspond to the watershed lines, which are the de- sired segmentation result, The result for this example is shown in Fig. 10.54(h) as dark, 1-pixel-thick paths superimposed on the original image Note the important property that the watershed lines form connected paths, thus giving continuous boundaries between regions ‘One of the principal applications of watershed segmentation is im the ex- traction of nearly uniform (bloblike) objects from the background, Regions characterized by small variations in intensity have small gradient values. Thus, in practice, we often see watershed segmentation applied to the gradient of an image, rather than to the image itself. In this formulation, the regional minima ‘of catchment basins correlate nicely with the small value of the gradient corre- ‘sponding to the objects of interest 105.2 Dam Construction ~ Before proceeding, let us consider how (o construct the dams or watershed lines required by watershed segmentation algorithms Dam construction is based on binary images, which are members of 2-D integer space 2” (see Section 2.4.2). The simplest way to construct dams separating sets of binary points is to use morphological dilation (see Section 9.2.2) The basics of how to construct dams using dilation are illustrated in Fig. 10.55. Figure 10.55(a) shows portions of two catchment basins at flooding step = ~ 1 ‘and Fig. 10.55(b) shows the result at the next flooding step, 1. The water has spilled from one basin to the other and, therefore, a dam must be built to keep ‘this from happening. In order to be consistent with notation to be introduced shortly. let M, and M; denote the sets of coordinates of points in two regional minima Then let the set of coordinates of points in the caichment basin associ- ‘ated with these two minima at stage n ~ 1 of flooding be denoted by C,—,(M,) ively. These are the two gray regions in Fig. 10.SS(a). denote the union of these two sets. There are two connected 10.55(a) (see Section 2.5.2 regarding connected compo- in Fig, 10.55(b). This counecteda b © da stage ~ | f hooding: HGURE 10.55 (a) Two partially flooded catchment basins at (b) Flooding at aft showing that swater has spitied PeTwee” basins. (¢) Strackunine Clement used for dilation. id) Result of dilation aia dam onstructioneae uring dilation, and condition (2) did not apply to any point during the dila- tion process; thus the boundary of each region was expanded uniformly. In the second dilation (shown in black), several points failed condition (3) while meeting condition (2), resulting in the broken perimeter shown in the fig- ture. It also is evident that the only points in q that satisfy the two conditions under consideration describe the I-pixel-thick connected path shown cromed- hatched in Fig. 10.55(d). This path constitutes the desired separating dam at stage n of Mlooding. Construction of the dam at this level of flooding is complet- ed by setting all the points in the path just determined to a value greater than the maximum intensity value of the image. The height of all dams is generally set at 1 plus the maximum allowed value in the image. This will prevent water from crossing over the part of the completed dam as the level of flooding is increased. It isimportant to note that dams built by this procedure, which are the desired segmentation boundaries, are connected components In other words, this method eliminates the problems.of broken segmentation lines. Although the procedure just described is based on a simple example, the method used for more complex situations is exactly the same, including the use ‘of the 3 x 3symmetric structuring element shown in Fig 10.55(c) 10.5.3 Watershed Segmentation Algorithm Let My, Mzy...,Ma be sets denoting the coordinates of the points in the sonal minima of an image g(x, »). As indicated at the end of Section 10.5.1, this typically will be a gradient image. Let C(§M)) be a set denoting the coordi- fates of the points in the catchment basin associated with regional minimem 1M {recall that the points in any catchment basin form a connected component). The notation min and max will be used to denote the minimum and maximum values of g(x, »). Finally, let T[n] represent the set of coordinates (s,) for which g(s, 0) < 7. That is, T[n] = {(s, gts. < nj (105-1) coordinates of points in g(x y) lying below Geometrically, T{n] is the set of the plane g(x, ¥) = = ; aa topography will be flooded in integer Mood increments ” eae + lton = max + 4. At any step 7 of the flooding process, the algo- ‘know the number of jints below the flood depth. ‘Conceptual- below the plane g(x. ») =" ‘marked white. Then when of flooding, we will see & ppoints in the function that quite useful in helping16.5 & Segmentation Using Morph ees (105-2) CAM) = CCM) OT) In other words, Cy(M,) = 1 at location (x,y) if (rye C(M)) AND (, y) « Tn}; otherwise C,(M) ‘ometrical interpretation of this fe- ‘ull is straightforward, We are simply using the AND operator to isolate at Stage n of flooding the portion of the binary image in Tr] that is associated with regional minimum M;, Next, we let C{] denotesthe union of the flooded catchment basing at stage . ln) = cot 05-3) Then Climax + 1] is the union of all catchment basins: tmx +1) = Uc) mt ‘Itcan be shown (Problem 10.41) that the elements in both C,(M,) and T(n) are never, replaced during execution of the algorithm, and that the number of ele- ments in these two sets either increases or remains the same as n increases Thus it follows that C[n ~ 1] is a subset of C[n}. According to Eqs. (10.5-2) and (105-3), Cf] is a subset of 7{n), so it follows that C[n ~ 1] is a subset of T[n), From this we have the important resull that each connected component ‘of Cjn — 1) is contained in exactly one connected component of T[} The algorithm for finding the watershed lines is initialized with Cfmnin + 1] = T[min + 1]. The algorithm then proceeds recursively, computing Cin] from C[n — 1]. A procedure for obtaining C[n] from Cn ~ 1] is as fol- jows. Let Q denote the set of connected components in T[n]. Then, for each connected component q € QJ}. there are three possibilities (105-4) 1. gAC[n - 1Jis empty. 2 gMC{n ~ 1} contains one connected component of C{n ~ 3] % gMCjn — 1} contains more than one connected component of Clr ~ 1}. Construction of C[n] from C[n — 1] depends off which of these three conditions holds Condition 1 occurs when a new minimum is encountered, in which case connected gis incorporated into C[n - 1] to form C[n]. Condition 2 ‘occurs when q lies within the catchment basin of some regional minimum, in which case q is incorporated into C[n ~ 1) to form C{n}. Condition 3 occurs when all, or part, of a ridge separating (wo or more catchment basins is en- countered. Further flooding would cause the water Jevel in these catchment basins to merge. Thus a dam (or dams if more than two catchment basins are involved) must be built within q to prevent overflow between the catchment basins. As explained in the previous section, a one-pixel-thick dam can be con- structed when needed by dilating g1C[n ~ 1] with a 3 x 3 structuring ele-