外文翻譯---改進(jìn)二維中值濾波器在線脈沖噪聲的抑制

1、<p>　　Improved 2-D Median Filter for On-Line Impulse Noise Suppressiom</p><p>　　Abstract-An inproved 2-D median filter employing multishell concept to suppress impulse noise ,is presented.The performanc

2、e of proposed filter is evaluated over image ‘LENA’ ,The impulsive noise is added using MATLAB utility.The modified strategy reduces the mnuber of replacement and results in better performance and simple hardware realiza

3、tion that is suitable for on-line implementation.</p><p>　　Index terms-Median Filter , Multi-shell Median Filter, Impulse Noise</p><p>　　I.INTRODUCTION</p><p>　　In TV and other ima

4、ging systems,impulse noise is a common impairment . The standard T.V.Broadcast signal is often contaminated with impulsive noise arising from various sources such as household electrical appliance and atmospheric disturb

5、ances.Broad banding of the signal further increases the level of impulsive noise. Various filters are proposed to suppress such impairments[1].The median filter(MF)[1-2] is widely used for impulse noise suppression and t

6、he multishell median filter(MMF)[3] intro</p><p>　　In this paper, the threshold strtegy of MMMF is modified so that:</p><p>　　effective noise filtering operations are performed under all conditi

7、ons,and</p><p>　　number of calculations/replacements is reduced and simplified.</p><p>　　This results in a simple hardware realization of the filter.</p><p>　　II.PROPOSED MODIFICATI

8、ON</p><p>　　Consider a 3x3-processing window, with P5 as the central pixel,as shown in Figure 1.</p><p>　　Fig.1. A 3x3 processing window </p><p>　　The output of MMMF as proposed i

9、n [4] is </p><p>　　Output (X,Y)= Max(P2,P8)if P5﹥Max[S]</p><p>　　P5 if Min [s]﹤Max[S]</p><p>　　Min(P2,P8) if P5﹤Max[S] (1)</p><p>　　Where S is the set of sam

10、ples surrounding central pixels except(P4.P6)i.e.</p><p>　　S=｛P1,P2,P3,P7,P8,P9｝ (2) </p><p>　　The principle invoved in the replacement strategy of Equation (1) is that if P5 is

11、corrupted by noise ,it is better to replace its gray level by P2 or P8 than by using Min[S] or Max[S] .also,due to missing lines error,since P4 and P6 may be lost, they are not considered in Equation(2).</p><p

12、>　　The limitation of Equation(1) is that when Min[S] or Max[S] are also corrupted by impulse noise,i.e.either Min[S] or Max[S] is equal to P5,Equation(1)fails to perform the desired filtering operation.To overcome thi

13、s limitation following modifications in the replacement strategy of Equation(1),are proposed.</p><p>　　Output (X,Y)= Max(P2,P8)if P5≥Max[S]</p><p>　　P5 if Min [s]＜P5＜Max[S]</p><p&g

14、t;　　Min(P2,P8) if P5≤Max[S] (3)</p><p>　　It has been observed that more than 70-80% points in an image,the gray level diatances of P5 from(P2 or P8) and from Max[S] are below 16.This is shown in Fig.2 for

15、 the image‘LENA’.This fact is used to further reduce unnessary replacements,thereby reducing the bluring of the images.Thus taking into considertion of Figure(3) can be further modified as</p><p>　　Output (

16、X,Y)= Max(P2,P8)if P5-Max[S]≥16</p><p>　　Max(P2,P8) if Min [s]-P5≥16</p><p>　　P5 otherwise （4）</p><p>　　Equation 4 indicates that replacing action takes place only whe

17、n the distance between P5 and Min[S] or Max[S] is no smaller than 16. This strtegy thus avoids the necessary replacements and reduces blurring of the images.Moreover， it can be implemented using simple comparators and su

18、btractors.</p><p>　　Gray level distances</p><p>　　Fig.2. Gray level distances between central point and its neighboring points for the image ‘LENNA’</p><p>　?、?.RESULTS</p>&

19、lt;p>　　Figure 3 shows the original image ‘LENNA’and Figure 4 shows the same image when corrupted with impulse noise. Results of median filter and the proposed filter are given in Figures 5 and 6, respectively.Comparin

20、g Figures 5 and 6, it is observed that the result of the proposed filter is much better than those obtained using the median filter. Aithough,the median filter remove the impulsive moise effectively, however,the image ge

21、ts blurred.The proposed filter removes the impulsive noise and also pr</p><p>　　A multishell filter employing the modified replacement strategy is presentde in this paper.The modified filter effectively supp

22、resses the inpulse moise.It uses threshold conditions that require fewer comparisons and replacements and is faster as compared to the other multishell median filters.moreover,it can be realized using simple comparators

23、and subtractors and subtractors and hence can be effectively used in real time applications </p><p>　　改進(jìn)二維中值濾波器在線脈沖噪聲的抑制</p><p>　　摘要：一種改進(jìn)二維中值濾波器，采用多殼的概念，以抑制脈沖噪聲，擬定的過(guò)濾器的性能進(jìn)行評(píng)估超過(guò)圖像“LENNA”的中值濾波，脈沖噪

24、聲被添加使用到MATLAB的實(shí)用工具中。修改后的策略減少了替換的數(shù)量，而且引出適合于上線實(shí)施更好的性能以及簡(jiǎn)單的硬件實(shí)現(xiàn)..</p><p>　　關(guān)鍵詞：指數(shù)計(jì)算中值濾波，多殼中值濾波，脈沖噪聲</p><p><b>　　引言</b></p><p>　　在電視和其他成像系統(tǒng)中，脈沖噪聲是一個(gè)共同的障礙。標(biāo)準(zhǔn)電視廣播信號(hào)往往被各種原因產(chǎn)生的脈

25、沖噪聲污染，如家電和大氣擾動(dòng)。加寬的信號(hào)進(jìn)一步提高脈沖噪聲水平。各種過(guò)濾器作出以抑制這種損傷[1]，中值濾波器（MF）[1-2]被廣泛用于脈沖噪聲抑制，多殼中值濾波器（MMF）[3]也介紹了失蹤線恢復(fù)的概念。盡管這些過(guò)濾器有令人滿意的表現(xiàn)，MMF的未能在相同的處理窗口，過(guò)濾兩個(gè)脈沖噪聲。此外，這些過(guò)濾器由于太多的替代趨于模糊的圖像。 C.J.Juan提出了改進(jìn)的多殼中值濾波器（MMMF）[4]。它消除大部分的MF和MMF的缺點(diǎn)。然而，它

26、得在一定條件下才能被使用，此將在后續(xù)的章節(jié)中討論。MMMF未能執(zhí)行所需的過(guò)濾操作。此外，在 MIN/ MAX的條件基礎(chǔ)上的計(jì)算/替代量仍然過(guò)大，使得過(guò)濾器，特別是實(shí)時(shí)應(yīng)用，難以實(shí)現(xiàn)。</p><p>　　在這個(gè)文件中，修改的MMMF限值策略使得：</p><p>　　（一）在任何情況下都進(jìn)行有效的噪聲過(guò)濾操作，</p><p>　　（二）減少和簡(jiǎn)化計(jì)算/替換。<

27、;/p><p>　　這樣就有了一個(gè)簡(jiǎn)單的過(guò)濾器的硬件實(shí)現(xiàn)。</p><p><b>　　擬議修改</b></p><p>　　擬定一個(gè)3x3的處理窗口，P5作為中央像素，如圖1所 </p><p>　　圖1 3x3的處理窗口</p><p>　　MMMF由[4]中輸出</p>&

28、lt;p>　　輸出（X,Y） = </p><p>　　這里S是除（P4.P6），圍繞中央像素的樣本集合</p><p>　　S=｛P1,P2,P3,P7,P8,P9｝ (2)</p><p>　　如果P5受噪聲污染，在方程（1）替代戰(zhàn)略所涉及的原則是，由P2或P8替代其灰度級(jí)比使用MIN[S]或Max[S]更好。同時(shí)，由

29、于失蹤線錯(cuò)誤，一旦P4和P6丟失，在方程（2）中他們將不會(huì)被考慮。方程（1）的限制，使得本來(lái)的Min[S] 或是Max[S]也會(huì)被脈沖噪聲損壞，要么是Min[S]要么是Max[S]等于P5，方程（1）未能執(zhí)行所需過(guò)濾操作。為了克服這一限制，接下來(lái)再對(duì)方程（1）替代策略進(jìn)行修改。</p><p>　　輸出 (X,Y)= </p><p>　　在圖像中，已觀察到70-80％以上的點(diǎn)，從（P2

30、或P8） 和Max[S]到P5的距離都低于16。這是圖2所示的圖像“LENNA”。 這實(shí)際上是用來(lái)進(jìn)一步減少不必要替代，從而降低圖像模糊。因此對(duì)圖（3）??可以進(jìn)一步修改。</p><p><b>　　輸出（X,Y）= </b></p><p>　　方程4表示，只有當(dāng)P5到Min[S] 或者 Max[S] 的距離不小于16采取替換措施。這一策略從而避免了必要替換并減少

31、圖像模糊。此外，它可以使用簡(jiǎn)單比較和減法。</p><p><b>　　灰度水平距離</b></p><p>　　圖2 中央點(diǎn)與其鄰近點(diǎn)灰度水平距離（圖像LENNA）</p><p>　　圖3 圖像LENNA檢查為100 (灰)點(diǎn)/英寸和8位/像素的掃描</p><p>　　圖4 圖像LENNA被脈沖莫伊茲損壞<

32、;/p><p>　　圖5 中值濾波器的輸出</p><p>　　圖6 擬定濾波器的輸出</p><p><b>　　3 結(jié)論</b></p><p>　　圖3顯示了原始圖像LENNA和圖4顯示了相同圖像被脈沖噪聲損壞時(shí)的圖像。在圖5和圖6分別給出中值濾波結(jié)果和擬議的濾波結(jié)果。比較圖5和6，發(fā)現(xiàn)擬議的濾波器結(jié)果比使用中值濾