201023055 六、發明辑明： 【發明所屬之技術領域】 本發明係為一種高效能指紋影像處理方法，其為一 種根據各點指紋流向，配合使用以具方向性之線型空間 .濾波器與菱型空間濾波器作為強化影像手段之指 . 像處理方法。 / 【先前技術】 人體之手掌及其手指、腳、峡时絲面具有凸凹不平的 ❿紋路’增加了皮膚表面的摩擦力，而包括指紋在内的這些皮膚紋 路在圖形、中斷點和交叉點之分布上每個人均不相同，具有其唯 -性’透過這種唯-性，每個人的驗可經過預先的採集與保留， 而作為-齡較哺證身分之科學猶手段，亦即為指紋 術。 议 土指紋識別技術的發展得益於現代電子集成製造技術和快逮可 #的演算法的研究’儘管指紋只是人體皮膚的—小部分，但用於 _識資料量卻相當大’由於指紋#料的比對上並非簡單的相等 (temple matching)與否，而是透過使用需進行大量運算的演 法，請配合參看第六圖所示，其中係現有技術中指紋識別 技術之步驟，而簡要言之，指紋識別技術主要涉及讀取指紋、 影像處理、提取特徵、資料提取及比對。 其中影像處理技術脚為先行通過指紋讀取贿讀取到 才曰紋的圖像’於取到指紋圖像之後，進一步對原始圖像進行 增強的處理，使之更清晰以得一低雜訊之高指紋影像品質， 而汾像處理後所得的影像品質則直接地影響了辨識 3 201023055 :在二^時=^^指：像缺點，往往 =線中央生成缺孔，可能於特徵萃取時被判斷為燥二 =取得的原始指紋圖像需做影像強化，修復缺孔、_等： 滤除雜訊，以降低辨識錯誤接收率(far)及錯誤拒絕率(=4 現有技術之麵指⑽像料技術 ）。 ❹ ::據影像頻率分布與方向進行__^ 盘、”、、而此作法通常耗費許多空間域與頻率域間的轉換時 …、法針對空間域的各點紋線方向不_進行個別增強，、:頻 ::全:方向的瓣再於空間域根據流向選擇正確增= 乍業上費時且又佔據大量之記憶體，· 另-個常用的灰階指纹縣增財朗為直接在空間域 衫象，例如非等向性類伽柏遽波器㈣沉咖anis〇 θ . filtering)、類局斯遮罩遽波器阳嶋㈤此刪 )= 波一 ~，但此些空_ 否二強’十分耗時’且通常需要佔用較大的遮罩空間， 疋设ΙΪ遇吊會退低於使用頻率域 驗即料絲賴爾敝^ _ 〜為解決上述之現有技術不足之處，本發明目的在提供一改良 之向效能缺影像處理方法歧善财技術巾影像處理費時且^ 費大量記憶體之缺失。 201023055 本發明係提供一種高效能指紋影像處理方法，係包 含下列步驟： 、隹而ί向判斷’其為於處理器上透過套人演算數值式， 的局判斷所採集之原始指紋灰階影像之特徵脊 Τ合濾波之線型濾波，其為於判斷出流向後的指紋 的，根據指紋線之紋向旋轉線型空間濾、波器並於處 ❹ ，套入線型遮罩渡波之定義式以進行各點之遽波； =其遮罩遽波作業，為利用係㈣χ3之據波進 其中定義式為： 〇〇〇〇〇〇〇〇〇. 1 1 〇 1 ο 1 〇 1 1 0 〇〇〇〇〇〇〇〇 2λ+12«+ι z:Z + ((W 一 (” + 切·cos θ(ζ·") + (V - 〇 +1)) sin θ〇· J)) / = y+(_ (Μ _ (w+1)}. sin θ{κ j}+(v_(m+1)} c〇s ^(. (昨，小坪，/)) 八 w;0 v=〇 、式中灰為線型遮罩的定義，A：為『遮罩係數和，彻力 為各點之流向角度，好(U)為該點之濾波結果； 複&濾波之菱形濾波，其為將經過線型濾波後之指 紋影，依據紋線方向旋轉菱型空間濾波器並於處理器曰 套入菱型ϋ罩瀘、波之定義式進㈣波演其遮罩遽波 作業，為利用係數9x3之濾波遮罩進行，其中定義式為： 5 201023055 0 0 4 l〇 14 l〇 4 〇 〇Ί W= 10 0 4 0 1 〇 4 〇 j〇 .° 〇 4 1〇 Μ w 4 〇 0j 2n+12m+l Κ = YJ'TdW{u,v) υ-0 ν=0 = / + ((U ~(η +1» · cos θ(ϊ, j) + (V - (W +1)} sin y)) 7=7 + (- (u - (η +1)) - sin θ(ΐ, j) + (ν _ (w +1)} c〇s ^ _ j 2μ+1 2m+\ Μ(ζ，= Ζ Σ Σ ν)χ Mil', /)) 、式申r為菱型濾波遮罩定義，尤為π遮罩係數和，外·刀 為各點之流向角度，D為該點之滤波結果； ^二值化，其為將經過處理器濾波處理增強後之指紋 y〆像進行—值化演算步驟，以顯示影像強化的效果。 而本發明具有下列優點： 1·本發明主要係結合線型遽波器具有高度連接性 及曼形遽波器具有可修補缺孔的優點，特別是指紋乾燥 部份能有效改善影像品質。 本毛明之複合濾 >皮器可針對各點不同流向進 影像增強，且使狀較鮮像數W9x3大小，藉 1降低㊉异像數可顯著地節錢波時間與記憶體空 …以降低指紋影像增強所耗#時間與儲存。< 【實施方式】 1 請配合參看第一圖所示，本發明之 :理方法為用於自動指紋辨識流程中； 用影像增強之複合遮罩演算法以處理所採集到之待 201023055 處理灰階指紋影像，並進_ 像’其可包含有下述步驟： 1.流向判斷 步將其轉換成二值牝之影 2·複合遽波之線型濾波 3 ·複合滤波之曼形遽波 4.二值化201023055 VI. SUMMARY OF THE INVENTION: [Technical Field] The present invention is a high-performance fingerprint image processing method, which is a linear space with a directionality according to the flow direction of fingerprints at each point. Filters and diamonds The spatial filter is used as a means of enhancing the imagery. / [Prior Art] The palm of the human body and its fingers, feet, and gorges have unevenly curved lines that increase the friction of the skin surface, and these skin lines including fingerprints are at the graphics, break points, and intersections. Each person in the distribution is different, with its uniqueness. Through this kind of uniqueness, each person's test can be pre-collected and retained, and as a scientific means of age-aged identity, it is Fingerprint surgery. The development of the fingerprint fingerprint recognition technology has benefited from the research of the modern electronic integrated manufacturing technology and the algorithm of the fast catching #. Although the fingerprint is only a small part of the human skin, the amount of data used for _ is quite large. The comparison of the materials is not simple (temple matching) or not, but through the use of a large number of operations, please refer to the sixth figure, which is a step in the prior art fingerprint recognition technology, and briefly In other words, fingerprint recognition technology mainly involves reading fingerprints, image processing, extraction features, data extraction and comparison. Among them, the image processing technology is the first to read the image of the fingerprint through the fingerprint reading. After the fingerprint image is taken, the original image is further enhanced to make it clearer to obtain a low noise. The high fingerprint image quality, and the image quality obtained after the image processing directly affects the recognition 3 201023055: in the case of ^^^^^: like the shortcomings, often = the center of the line generates the missing holes, which may be Judging as dry two = the original fingerprint image obtained needs to be image-enhanced, repairing the missing hole, _, etc.: filtering out the noise to reduce the recognition error rate (far) and false rejection rate (=4 the face of the prior art (10) Image technology). ❹ :: According to the image frequency distribution and direction, __^ disk, ",, and this method usually consumes a lot of conversion between the spatial domain and the frequency domain..., the method does not _ individual enhancement for the spatial direction of the dot ,::Frequency::All: The direction of the flap is then increased in the spatial domain according to the flow direction. 乍 It takes time and takes up a lot of memory. · Another common grayscale fingerprint county is rich in space. Domain shirts, such as non-isotropic Gabor chopper (four) Shen coffee anis 〇 θ. filtering), class stalk mask chopper Yang (5) this delete) = wave one ~, but this empty _ no The second strong 'very time consuming' and usually need to occupy a large mask space, the ΙΪ ΙΪ 会 会 会 会 会 低于 使用 使用 使用 使用 使用 赖 赖 为 为 为 为 为 为 为 为 为 为 为 为The object of the present invention is to provide an improved image processing method for cost-effective image processing, which is time-consuming and requires a large amount of memory. 201023055 The present invention provides a high-performance fingerprint image processing method, which comprises the following steps:隹 ί ί to judge 'it is on the processor After the set of human calculus numerical value, the local judgment of the original fingerprint grayscale image is collected by the characteristic ridge filtering and filtering linear filtering, which is to determine the flow of the backward fingerprint, according to the fingerprint line to the rotating linear space filter, The wave is placed at the same place, and the definition of the line-shaped mask is applied to perform the chopping of each point; = the masking operation of the line is used to make the use of the system (4) χ3 into the definition: 〇〇〇〇〇 〇〇〇〇. 1 1 〇1 ο 1 〇1 1 0 〇〇〇〇〇〇〇〇2λ+12«+ι z:Z + ((W one (" + cut·cos θ(ζ·") + (V - 〇+1)) sin θ〇· J)) / = y+(_ (Μ _ (w+1)}. sin θ{κ j}+(v_(m+1)} c〇s ^ (. (Yesterday, Xiaoping, /)) Eight w; 0 v = 〇, the definition of the gray is the linear mask, A: is the "mask coefficient and, the force is the flow angle of each point, good (U) The filtering result for this point; the diamond filtering of the complex & filtering, which is a fingerprint shadow after linear filtering, rotating the diamond spatial filter according to the direction of the ridge, and inserting the diamond ϋ 泸, wave in the processor 曰The definition of the (four) wave performs its mask chopping operation, in order to utilize the coefficient 9x3 filter mask is performed, where the definition is: 5 201023055 0 0 4 l〇14 l〇4 〇〇Ί W= 10 0 4 0 1 〇4 〇j〇.° 〇4 1〇Μ w 4 〇0j 2n +12m+l Κ = YJ'TdW{u,v) υ-0 ν=0 = / + ((U ~(η +1» · cos θ(ϊ, j) + (V - (W +1)} Sin y)) 7=7 + (- (u - (η +1)) - sin θ(ΐ, j) + (ν _ (w +1)} c〇s ^ _ j 2μ+1 2m+\ Μ( ζ, = Ζ Σ Σ ν) χ Mil', /)), the formula is r is the definition of the diamond filter mask, especially the π mask coefficient and the outer knife is the flow angle of each point, D is the filtering of the point Result; ^ Binarization, which is a step of performing a value-enhancing calculation on the fingerprint y image enhanced by the processor filtering process to display the effect of image enhancement. The present invention has the following advantages: 1. The present invention mainly relates to the high connectivity of the linear chopper and the omni-directional chopper having the advantage of repairing the missing holes, in particular, the fingerprint drying portion can effectively improve the image quality. Ben Maoming's composite filter > leather device can increase the image direction for different points, and make the shape of the fresh image number W9x3. By reducing the number of ten images, the money can be significantly reduced by the time and memory. The time and storage of fingerprint image enhancement. <Fifth Embodiment 1 Please refer to the first figure, the method of the present invention is used in the automatic fingerprint identification process; the image-enhanced composite mask algorithm is used to process the collected processing time 201023055 The order fingerprint image, and the _ image can contain the following steps: 1. Flow direction judgment step converts it into binary 牝 shadow 2 · Composite chord line type filter 3 · Composite filter 曼 shape chopping wave 4. Value
别述之流向判斷亦即為判斷原始指、纹流向，於處理 器執行影像增強之演算前需先評估所採集之指紋的局 部走向，以便於取得之指紋流向上套㈣波遮罩進行後 續演算’而於指紋各點上取得減流向特徵之方法已為 一熟知之技術’本說明書利用於處理器中套用史塔克及史 _(Stock and Sw〇nger)所提出之指紋流向判別方法為 實施例’其數值式定義如下： 50 = M(i, j+ 4) +M (i, J+ 2) + M(i, y-2) + M(i, j - 4)The flow direction judgment is to judge the original finger and the flow direction. Before the processor performs the image enhancement calculation, it is necessary to first evaluate the local trend of the collected fingerprint, so as to obtain the fingerprint flow upward (four) wave mask for subsequent calculation. 'The method of obtaining the current-reducing feature at each point of the fingerprint is a well-known technology. This specification uses the fingerprint flow direction discrimination method proposed by Stock and Sw〇nger in the processor. Example 'The numerical formula is defined as follows: 50 = M(i, j+ 4) +M (i, J+ 2) + M(i, y-2) + M(i, j - 4)
Sx =^0-2,y+ 4) + M(/-l,y+ 2) +Af(/ + i；y_2) + ^(/ + 2jy_4) ^=^(/-4,/ + 4) + ^(/-2,7 + 2) + ^(/ + 2,/-2) + ^(/ + 4,7-4) =M(/-4,7·4-2) + 71/(ζ-2,7 + 1) + 71/〇· + 2)7·-1) + ^(/ + 4,y-2) = M(i - 4, j) + M(i- 2, j) + M{i + 2, j) + M(i + 4, j) (A)Sx =^0-2,y+ 4) + M(/-l,y+ 2) +Af(/ + i;y_2) + ^(/ + 2jy_4) ^=^(/-4,/ + 4) + ^ (/-2,7 + 2) + ^(/ + 2,/-2) + ^(/ + 4,7-4) =M(/-4,7·4-2) + 71/(ζ- 2,7 + 1) + 71/〇· + 2)7·-1) + ^(/ + 4,y-2) = M(i - 4, j) + M(i- 2, j) + M {i + 2, j) + M(i + 4, j) (A)
Ss=M(i-4,J~2) + M(i-2,j-l) + M(i + 2,j + l) + M(i+4,j + 2) S^m-4,j-4) + M(i-2,j-2) + M(i + 2,j+2) + M(i + 4,j + 4) S, =^(i-2,j-4) + M(i-l,J-2) + M(i + l,j + 2) + M(i + 2,j+4) ⑷式中鄉·，_/)為指紋圖像M於各點處之灰度值，而AS K ^係 各個方向之灰度值總和，而&ax和5'πώ分別表示各方向中最大與最小 灰度值和，其如(Β)式所示_· ^raax = max sn η = Ο,Κ ,7 (Β) «*0^,7 ηSs=M(i-4,J~2) + M(i-2,jl) + M(i + 2,j + l) + M(i+4,j + 2) S^m-4,j -4) + M(i-2,j-2) + M(i + 2,j+2) + M(i + 4,j + 4) S, =^(i-2,j-4) + M(il,J-2) + M(i + l,j + 2) + M(i + 2,j+4) (4) where the township, _/) is the gray of the fingerprint image M at each point Degree, and AS K ^ is the sum of the gray values in all directions, and &ax and 5'πώ respectively represent the sum of the maximum and minimum gray values in each direction, as shown by (Β) _· ^raax = Max sn η = Ο,Κ ,7 (Β) «*0^,7 η
SmiB = min Si η = Ο,Κ ,7 «=η ν η 1 9 5 7 201023055 而根據上述之計算結果可判斷各點處之指紋流向特徵值 者，令㈣)為術)點之流向角度，套用如下列(Q所示，根據㈣ 點之埘以）、，判斷正確方向& ’ ， 度㈣： 更了。十异出Μ點處之角 州·场·，力 (C)SmiB = min Si η = Ο, Κ , 7 «= η ν η 1 9 5 7 201023055 And according to the above calculation results, the fingerprint flow at each point can be judged to be the characteristic value, and (4)) is the flow angle of the point) Apply as follows (Q, according to (4) points), determine the correct direction & ', degree (four): more. The corner of the ten different points of the state, the field, the force (C)
Amin otherwise 紙/) = 4.22.5° 而透過於處理器中取得各點之 局部之指纹、户θ4θ 角度後即可知悉各 巧I H日、.文抓向，如第二圖所示 向#徵夕卷M s M + A $ 4才曰、蚊原圖取出流 ㈣被之含的局σρ方向，然而於處理 以進行圖像數值之演算已為—孰 用禮值式 未就其間轉換之細節於說明書中多4述二套： 運算之計算式作說明。 僅就其套用 前述之複合遽波之線型濾波步驟為 e 後的指紋影像上，根m纟^ # ^ ^ -- 疋得線型空間濾波器並 於慝理斋進订濾波演算，其中線 波如第二m所千.主i間濾波器之遮罩濾 及如弟一圖所不，在本實施例為 遮罩於夂赴，隹—机任 』用係數為9x3之濾波 侉體处門，而甘A細， ^即’濾波時間及記 11體工間，而其中線型遮罩濾波之 〇〇〇00〇〇〇〇 0 1 0 Γο Ο π η η η Λ „ ^ @ Α 如下： (D) 2n+l2m+l ^=ΣΣ^(^ν) (Ε) 8 201023055 ϊ = i + ((« - (« +1)). cos 9{i, j) + {v~(m +1)) sin 9{U j)) ·/ = 7 + (— (w — +1)). sin θ(ζ·，/) + (v _ (tw +1)) cos 0(z*，/)) (F) _ 1 2n+\2m+\ 其中（D)式之π為線型遮罩的定義，⑹式中之尤為『 遮罩係數和，而藉由（F)式、（G)式可根據流向判斷中之 /;IL向角度叩，/)而後運异出濾'波結果_，/) ’而因為線型遽 波步驟中使濾波因素加入指紋流向，透過該加入流向因 ❹ 素以達到指紋斷線修補並避免平行紋線相接等目的，對 於紋線之斷裂有良好的修復效果，運算量低且簡便。 前述之複合濾波之菱型濾波步驟為：將經過線型濾 波後之指紋影像依據紋線方向旋轉菱型空間濾波器並 於處理器進行濾波演算，其中菱型空間濾波器之遮罩濾 波處理時，亦為利用像數為9x3之濾波遮罩，而後於各 點進行數值之判讀轉換以於紋線的垂直方向上提高脊 線與谷線之對比度’而其中菱型遮罩濾波處理之定義式 ⑩如下： 0 0 4 10 14 10 4 0 0 w = 10 0 4 0 1 0 4 0 10 0 0 4 10 14 10 4 0 0Amin otherwise paper /) = 4.22.5 ° and through the processor to obtain the local fingerprint of each point, the angle of θ4θ, you can know the IH day, the text capture, as shown in the second figure to #征夕卷M s M + A $ 4 曰, the mosquito original map takes out the stream (4) is contained in the direction of the σρ, but the processing for the calculation of the image value is already - the details of the ritual value is not converted In the manual, there are four sets of two sets: The calculation formula of the operation is explained. Only on the fingerprint image after the linear filtering step of the composite chopping described above is applied, the root m纟^ # ^ ^ -- 疋 线 型 空间 空间 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线 线The second m is thousands. The mask filtering of the filter between the main i and the other is not shown in the figure. In this embodiment, the mask is placed on the 夂, and the machine is equipped with a filter body with a coefficient of 9x3. And Gan A is fine, ^ is the 'filtering time and remembers 11 bodywork, and the linear mask filtering 〇〇〇00〇〇〇〇0 1 0 Γο Ο π η η η Λ „ ^ @ Α is as follows: (D 2n+l2m+l ^=ΣΣ^(^ν) (Ε) 8 201023055 ϊ = i + ((« - (« +1)). cos 9{i, j) + {v~(m +1) ) sin 9{U j)) ·/ = 7 + (— (w — +1)). sin θ(ζ·,/) + (v _ (tw +1)) cos 0(z*,/)) (F) _ 1 2n+\2m+\ where π of (D) is the definition of linear mask, and (6) is especially the mask coefficient and, by (F) and (G), it can be judged according to the flow direction. In the / / IL direction angle 叩, /) and then transport out the filter 'wave results _, /) ' and because the line chopping step in the filter factor into the fingerprint flow, through the join flow direction In order to achieve the purpose of fingerprint breakage repair and avoiding the parallel lines, it has a good repair effect on the break of the line, and the calculation amount is low and simple. The above-mentioned composite filtering diamond filtering step is: after linear filtering The fingerprint image rotates the diamond spatial filter according to the direction of the line and performs filtering calculation on the processor. In the mask filtering process of the diamond spatial filter, the filter mask with the image number of 9x3 is used, and then the points are used. The numerical interpretation is performed to increase the contrast between the ridge line and the valley line in the vertical direction of the line', and the definition of the diamond mask filtering process is as follows: 0 0 4 10 14 10 4 0 0 w = 10 0 4 0 1 0 4 0 10 0 0 4 10 14 10 4 0 0
2«+12m+I (H) (I) (J) (K) ι/κ：〇 ν=0 i'= ^ + (iu ~(n +1)) · cos θ(ΐ, j) + (v-(m +1)) sin θ(ϊ, j)) j' = j + {~(u~(n +1)) · sin θ(ΐ, j) + (v - (w +1)) cos 9{i, j))2«+12m+I (H) (I) (J) (K) ι/κ:〇ν=0 i'= ^ + (iu ~(n +1)) · cos θ(ΐ, j) + ( V-(m +1)) sin θ(ϊ, j)) j' = j + {~(u~(n +1)) · sin θ(ΐ, j) + (v - (w +1)) Cos 9{i, j))
_ 1 2w+12fti+I = — Σ 2(^(w»v)XjWO-，=/)) t/s〇 Vs=〇 9 201023055 /⑻式h為菱型濾波遮罩定義，（1)式中之以㈣罩 係數和而藉由⑺式、(κ)式可根據流向判斷中之流向角 度㈣而後運算出遽波結，而因為菱型渡波步驟 /中使慮波因素加人指紋流向，而可達到修復破碎且滤除 . 谷線汙損，以提升脊線與谷線間區別度之功效，請進一 •參f第四圖所示，其中為指紋原圖經過複合濾波 作業後所得之影像增強實施成果圖。 ❹ 刖述之二值化步驟為將經過處理器濾波處理增強 後之指紋影像進行二值化演算步驟，即可明顯看出影像 強化的效果’以便將影像作後續之比對。以下為使用動 態閥值之二值化演算法，數值式如下式所示： j) = Σ Σ ^〇 + j + ^)] - (2n + if /r λ . ^ V,-*—/y M{i,j)J255 Hij)>T{i,j) l〇 Jf (M) 目動態閾值二值化不使用單-二值化閥值，故可降 低按壓不均勻衫響，（L)式之印,力為根據⑹)點及其周圍 點灰度值分佈所訂出的閥值，（M)式則根據各點閥值 叩’力來决疋一值化結果泣(/，力，而因為二值化步驟為一現 有技術中熟知之特徵汲取手段，因此亦不於說明書中對 其細節多加探討。 本發明之主要特徵為將指紋灰階影像之斷線連接 與雜訊遽除分成兩個步驟：線型滤波、菱形遽波來完 成’此兩種具方向性之複合濾波方式使用遽波器之運算 201023055 簡早’可達到良好的影像增強效果，避免 且結合線型濾波器具有高度連接性、1特徵點， 補缺孔的優點作為強化影像 慮波器可修 法，可有效地修補非自然原因之紋又線斷;致二像處理方 缺陷，例如按壓手指乾燥等缺陷指紋，二Γ =紋空洞等 看第五圖所示，其為隨機選取 ：進-步配合參 Ο 魯 統之頻域遽波作業_及利用進行傳 (C)在二值化没取特徵之後的顯像差星7〜慮波作業 清楚比照出本發明作業所產生之影像優勢自其中可 再者’本發明使用的濾波遮罩像數理 小，但實際設計時可使用小於9X3❺遮罩'4 9x3大 地節省渡波時間與記憶體空㈤，以降低’ 2此可顯著 耗費時間與儲存空間。 _ 9、、文影像增強所 综觀上述，本發明在突破先前之技術下 進之功效’且也非熟悉該項娜者所易於思及 到所欲增 請前未曾公開，其所具之進步性、實· ’本發明申 申請要件，贿出創作申請，:Γ貴=發明專利之 申請案，以勵創作，至感德便。 、。"隹本件創作專利 以上所述之實施例僅係為說明本發明 ^錢嶋娜之人地吻本發^ 施，g不能以之限定本發明之專利範園，即大凡依本發明所揭貧 之^所作之騎變化飾，域_在本㈣ ^ 【圖式簡單綱】 觀_。 第一圖係為本發明影像處理流程圖。 201023055 第二圖係為本發明之指紋流向演算結果對照圖。 第三圖係為本發明遮罩濾波示意圖。 第四圖係為本發明影像增強並二值化結果對照圖。 第五圖係為本發明之影像增強與現有技術之影像處理。 第六圖係為指紋自動辨識系統程序流程圖。 【主要元件符號說明】 (a)指紋原圖 (b)頻域濾波作業 (c)複合濾波作業 12_ 1 2w+12fti+I = — Σ 2(^(w»v)XjWO-,=/)) t/s〇Vs=〇9 201023055 /(8) where h is the definition of the diamond filter mask, (1) The (4) cover factor and the (7) and (κ) formulas can be used to calculate the chopping wave according to the flow direction angle (4) in the flow direction judgment, and because the diamond wave step/middle causes the wave factor to add the fingerprint flow direction, It can achieve the effect of repairing broken and filtering. Valley line pollution, in order to improve the difference between ridge line and valley line, please enter the first picture of Figure 4, which is the result of the original fingerprint image after the composite filtering operation. Image enhancement implementation results map.二 The binarization step of the description is to perform the binarization calculation step on the fingerprint image enhanced by the processor filtering process, so that the effect of image enhancement can be clearly seen' for subsequent comparison of the images. The following is a binary algorithm using dynamic thresholds. The numerical expression is as follows: j) = Σ Σ ^〇+ j + ^)] - (2n + if /r λ . ^ V,-*-/y M{i,j)J255 Hij)>T{i,j) l〇Jf (M) The dynamic threshold binarization does not use the single-binarization threshold, so the uneven press of the press can be reduced. ), the force is based on (6)) the threshold value of the point and its surrounding point gray value distribution, (M) is based on the threshold value of each point to determine the value of the result of the cry (/, Force, and because the binarization step is a well-known feature extraction method in the prior art, it is not discussed in detail in the specification. The main feature of the present invention is the disconnection of the grayscale image of the fingerprint and the noise. In addition to two steps: linear filtering, diamond-shaped chopping to complete 'the two directional composite filtering method using chopper operation 201023055 simple early' can achieve good image enhancement, avoid and combine linear filter with High connectivity, 1 feature point, and the advantage of filling the hole as a reinforcement image filter can be repaired, which can effectively repair the lines and lines of unnatural causes. Broken; caused by two image processing defects, such as pressing fingerprints such as finger drying, etc., see the fifth picture, which is shown in the fifth figure, which is randomly selected: the step-step cooperation with the reference Ο Lutong's frequency domain chopping operation _ And the use of the transmission (C) after the binarization of the feature is not taken out of the star 7 - wave operation clearly compared to the image superiority produced by the operation of the present invention from which the filter mask image used in the present invention The mathematics is small, but the actual design can use less than 9X3 ❺ mask '4 9x3 to save the wave time and memory space (5), to reduce '2 this can significantly consume time and storage space. _ 9, text image enhancement is comprehensive The invention has the effect of breaking through the previous technology's and is also unfamiliar with the person who is familiar with the item, and it is easy to think that the desired increase has not been disclosed before, and the progress and the actual application requirements of the present invention are Bribery application for creation: Γ贵 = application for invention patent, to encourage creation, to the sense of virtue. 。 。 隹 隹 创作 创作 创作 创作 创作 创作 创作 创作 创作 创作 创作 创作 创作 创作 创作 创作 创作 创作 创作 创作 创作 创作 创作 创作 创作 创作 创作 创作 创作 创作Kiss this hair ^, can not be used The patent garden of the present invention, that is, the riding variation made by the general inventors of the present invention, is in the form of the image processing flowchart of the present invention. 201023055 The second figure is a comparison diagram of the fingerprint flow calculation results of the present invention. The third figure is a schematic diagram of the mask filtering of the present invention. The fourth figure is a comparison diagram of the image enhancement and binarization results of the present invention. The image enhancement of the present invention and the image processing of the prior art. The sixth figure is a program flow chart of the automatic fingerprint identification system. [Explanation of main component symbols] (a) Original fingerprint (b) Frequency domain filtering operation (c) Composite filtering operation 12