2023年全國碩士研究生考試考研英語一試題真題(含答案詳解+作文范文)_第1頁
已閱讀1頁,還剩19頁未讀 繼續(xù)免費(fèi)閱讀

下載本文檔

版權(quán)說明:本文檔由用戶提供并上傳,收益歸屬內(nèi)容提供方,若內(nèi)容存在侵權(quán),請進(jìn)行舉報(bào)或認(rèn)領(lǐng)

文檔簡介

1、<p>  南京郵電大學(xué)通達(dá)學(xué)院</p><p>  畢業(yè)設(shè)計(jì)(論文)外文資料翻譯</p><p>  學(xué) 院: 南京郵電大學(xué)通達(dá)學(xué)院 </p><p>  ?! I(yè):    軟件工程    </p><p>  學(xué)生姓名:    張 峰     </p><p>  班級學(xué)號:    08003

2、019    </p><p>  外文出處:  《物聯(lián)網(wǎng)技術(shù)》   </p><p>  附件:1.外文資料翻譯譯文;2.外文原文 </p><p><b>  復(fù)雜脊波圖像去噪</b></p><p>  作者:G. Y. Chen and B. Kegl ,刊名:Pattern Recognition,出

3、版日期:2007</p><p><b>  摘要</b></p><p>  脊波變換是在小波變換的基礎(chǔ)上提出的多尺度分析方法,對于圖像中直線狀和超平面的奇異性問題,脊波變換比小波變換有更好的處理效果,應(yīng)用數(shù)字復(fù)合脊波變換去除嵌入在圖像中的白噪聲,并使用一個(gè)簡單的復(fù)合脊波系數(shù)的硬閾值來實(shí)現(xiàn),實(shí)驗(yàn)結(jié)果表明,種算法比VisuShrink算法、普通脊波算法和Wiener2

4、濾波器圖像去噪的去噪效果更好,同時(shí)復(fù)合脊波算法也能應(yīng)用于圖像去噪和模式識別特征提取。</p><p>  關(guān)鍵詞:圖像去噪;小波變換;脊波變換;復(fù)合脊波</p><p><b>  1.介紹</b></p><p>  小波變換已成功應(yīng)用到許多科學(xué)領(lǐng)域,如:圖像的壓縮、圖像去噪、信號的處理、計(jì)算機(jī)繪圖和模式識別等等。但小波變換對于奇異性問題,如

5、數(shù)字圖像中的邊界</p><p>  以及線狀特征等,不是非常有效。這是基于小波的處理方法,如圖像壓縮和去噪 等應(yīng)用中,不可避免地在圖像邊緣和細(xì)節(jié)上有一定程度的模糊,然而這些不連續(xù)</p><p>  特征恰恰可能是信號最重要的信息。因此,Donoho等在小波變換的理論基礎(chǔ)上建立了一種適合表示奇異性的多尺度方法,這種方法稱為脊波變換。脊波是在小波變換基礎(chǔ)添加了一個(gè)表征方向參數(shù)得到的,因此,

6、它與小波一樣也具有局部時(shí)頻分辨能力,同時(shí)還具有很強(qiáng)的方向選擇和辨識能力,能非常有效表示信號中具 方向性的奇異特征。實(shí)驗(yàn)表明脊波在直線特征的表示和提取中非常有效。</p><p>  經(jīng)過多年的發(fā)展,脊波變換打破了小波變換的局限性,二維小波變換圖像可生成大的小波系數(shù)并在每個(gè)尺度上進(jìn)行分解。因?yàn)樵谌绱舜蟮男〔ù笙禂?shù)下,采用小波更換噪聲圖像去噪面臨著許多困難。目前,脊波變換已成功地應(yīng)用到數(shù)字 圖像分析,與小波變換不同的

7、是,脊波變換是在各方向奇異性的取向和定位的積分式變換。脊波是常數(shù),其方程式為x1 cosØ+x2cos Ø=c。其中,c為常數(shù),在這些脊波方向上的正交處正好是小波系數(shù)。在脊波變換中結(jié)合了二元樹復(fù)合小波變換,并把它應(yīng)用于圖像去噪。實(shí)驗(yàn)結(jié)果表明,采用二元樹復(fù)合脊波算法能獲得比其他圖像去噪算法更高的峰值信噪比。</p><p>  這篇文章大體是這樣的。在第二部分,我們將解釋如何將二元樹復(fù)雜的波變換

8、成脊波圖像去噪。實(shí)驗(yàn)結(jié)果在第3節(jié)。第4節(jié)是最后得出的結(jié)論和未來需要做的工作。</p><p>  2.用復(fù)雜脊波圖像去噪</p><p>  離散脊波變換提供了兩個(gè)光滑物體和物體邊緣的稀疏性近乎理想的描述,它 是高斯噪聲去噪接近于理想方法。脊波系數(shù)較小的數(shù)字脊波變換可以壓縮圖像的能量,另一方面,小波變換是分解每一個(gè)二維小波尺度,從而在圖像邊緣處產(chǎn)生許多大的小波系數(shù),這意味著許多小波系數(shù)必須

9、重新構(gòu)建。</p><p>  以數(shù)字?jǐn)?shù)據(jù)為近似的Radon變換是基于離散快速傅里葉變換的。普通脊波變換能夠?qū)崿F(xiàn)如下功能:</p><p>  (1)計(jì)算圖像二維快速傅立葉變換(FFT);</p><p> ?。?)用取樣值的極性方格替換傅里葉變換獲得方格取樣值;</p><p>  (3)計(jì)算一維角線的反向快速傅里葉變換;</p>

10、;<p> ?。?)執(zhí)行一維標(biāo)量小波變換所產(chǎn)生的角線以獲取脊波系數(shù)。</p><p>  普通的離散小波變換沒有平移不變性,當(dāng)信號輸入時(shí)一個(gè)小變化會導(dǎo)致不同的小波系數(shù),為了克服這個(gè)問題,Kingsbury引入了一種新型的小波變換方法, 該變換稱為二元樹復(fù)合小波變換,他闡明了近似的平移不變性和改善角分辨率。由于標(biāo)量小波沒有移位不變性,二元樹復(fù)合小波變換最好是運(yùn)用脊波變換,稱之為復(fù)合脊波變換。最后一步脊

11、波變換,能夠用一維二元樹復(fù)合小波變換代替一維標(biāo)量小波變換。這樣,脊波變換能較好地結(jié)合二元樹復(fù)合小波變換的移動不變性性能。 </p><p>  復(fù)合脊波變換可以應(yīng)用于整個(gè)圖像,也可以把圖像分割成若干個(gè)相互重疊正方格并且每個(gè)正方格運(yùn)用脊波變換中。分解n*n原圖像為平滑地邊長為R像素相</p><p>  互重疊塊,以致重疊區(qū)兩兩垂直方向鄰接塊之間是一個(gè)長度為R/2*R矩陣列,同時(shí)這個(gè)重

12、疊區(qū)兩兩平行方向鄰接塊之間是一個(gè)R*R/2矩陣列。對于一個(gè)n*n的圖</p><p>  像,期望每個(gè)方向?yàn)?n/R的塊,這種分割方法會產(chǎn)生4倍的冗余。為了獲得去噪復(fù)合脊波系數(shù),在當(dāng)前像素位置使用4個(gè)去噪復(fù)合脊波系數(shù)平均值。</p><p>  對于復(fù)合的脊波變換的閾值是類似于曲波閾值,區(qū)別在于,當(dāng)取復(fù)合脊波系數(shù)尺度的閾值時(shí),令yλ是噪聲脊波系數(shù)。用下面硬閾值定律來估計(jì)未知的脊波 系數(shù)。當(dāng)

13、|yλ|》kδδ時(shí),令yλ=yλ,否則yλ=0。其中,δ用近似Monte-Carlo模擬,常數(shù)k依賴于噪聲δ。當(dāng)噪聲δ小于30時(shí),設(shè)k=5為第一分解尺度并且設(shè)k=4為其他分解尺度。當(dāng)噪聲δ大于30時(shí),設(shè)k=6為第一分解尺度并且設(shè)k=5為其他分解尺度。</p><p>  復(fù)合脊波圖像去噪算法的描述如下:</p><p>  (1)把圖像分割成R*R區(qū)域,兩相鄰區(qū)域相互垂直重疊R/2*R個(gè)像

14、素,兩相鄰 區(qū)域水平重疊R*R/2個(gè)像素。</p><p>  (2)對于每個(gè)塊,基于應(yīng)用提出了復(fù)合脊波,閾值的復(fù)合脊波系數(shù),及進(jìn)行復(fù)合脊波逆變換。</p><p> ?。?)在同一點(diǎn)的圖像去噪的像素值取平均值。</p><p>  這種算法稱為復(fù)合脊波壓縮算法而該算法,使用普通脊波壓縮,通過使用標(biāo)量小波使復(fù)合脊波壓縮的計(jì)算復(fù)雜性的與脊波壓縮復(fù)雜性相當(dāng)。惟一差異是采

15、 用一維二元樹復(fù)合小波變換取代一維小波變換,在計(jì)算量上一維二元樹復(fù)合小波是一維標(biāo)量小波的兩倍。然而,該算法的其他算法步驟有同樣計(jì)算量實(shí)驗(yàn)結(jié)果表明復(fù)合脊波壓縮優(yōu)于VisuShrink算法、脊波壓縮算法和Wiener2的所有測試案例的濾波器。在某些情況下,獲得0.8dB峰值信噪比超過了脊波壓縮算法。圍繞</p><p>  VisuShrink更是為所有圖像去噪更大的改善,這表明復(fù)合脊波去噪算法是自然圖像去噪的最佳選

16、擇。</p><p><b>  3.實(shí)驗(yàn)結(jié)果</b></p><p>  在實(shí)驗(yàn)中使用著名的Lena圖像,在圖像中將不同噪聲級別的高斯白噪聲加入到原始無噪聲圖像產(chǎn)生噪聲圖像。對VisuShrink,RidgeletShrink,復(fù)合脊波去噪</p><p>  和Wiener 2濾波器進(jìn)行比對,VisuShrink是運(yùn)用普遍的軟閾值去噪技術(shù),

17、Wiener 2函數(shù)由Matlab圖像處理工具箱提供,使用圖像中每個(gè)像素5*5鄰域。wiener2函數(shù)適用于Wiener濾波器(線性濾波器的一種)的圖像自適應(yīng),剪裁圖像自身的局部圖像方差,信號的峰值信噪比(PSNR)的實(shí)驗(yàn)結(jié)果見表1。把圖像劃分為32*32或64*64的塊尺寸是最好的選擇,表1表明了圖像Lena去噪效果,在表格中的第</p><p>  一列是原始圖像噪聲的PSNR,而其他列都采用不同的去噪算法得

18、到去噪后圖像的峰值信噪比。PSNR定義如下:PSNR = 10 log10Pi;j (B(i; j) A(j))2n22552 :;</p><p>  式中:B為有噪聲圖像,A為無噪聲的圖像。從表1可看到,復(fù)合脊波去噪算法優(yōu)于VisuShrink,普通脊波去噪和Wiener 2。當(dāng)噪聲級別低的時(shí)侯,VisuShrink</p><p>  無任何去噪能力,在這種情況下,VisuShrin

19、k的去噪甚至比原噪聲圖像更差的圖像效果。然而,在這種情況下復(fù)合脊波去噪效果相當(dāng)不錯(cuò)。對于某些情況下,復(fù)合脊波與普通脊波去噪相比能夠得到約0.8dB的改善。這表明,通過二元樹復(fù)合小波結(jié)合脊波變換能得到圖像去噪意義的改善,復(fù)合脊波算法比VisuShrink算</p><p>  法對圖像的去噪效果更好,甚至更有意義的是在所有噪聲級別和測試圖像。圖1顯示了無噪聲的原始圖像,圖像噪聲增加,VisuShrink去噪圖像,普

20、通脊波圖像 去噪,復(fù)合脊波去噪圖像和Wiener 2圖像去噪處理后的Lena圖像,以上實(shí)驗(yàn)都是在32*32像素劃分塊尺度的條件下進(jìn)行的。因此,就直線性和曲線的特征和高質(zhì)量的邊緣恢復(fù)方面言,復(fù)合脊波去噪產(chǎn)生的視覺更清晰的圖像降噪效果比VisuShrink、普通脊波去噪和Wiener 2濾波器都更好。 </p><p><b>  4.結(jié)論和未來工作</b></p><p&

21、gt;  研究使用復(fù)合脊波的圖像去噪方法。復(fù)合脊波變換是通過一維二元樹復(fù)合小波變換轉(zhuǎn)換到Radon變換系數(shù)獲得。在近似平移的二元樹復(fù)合小波變換不變性,從而使用復(fù)合脊波變換的圖像去噪一個(gè)很好的選擇。復(fù)合脊波變換能提供光滑物體和物體邊緣接近理想稀疏性,這使得噪聲脊波閾值系數(shù)的高斯白噪聲去噪接近</p><p>  最佳方法。為測試新的去噪方法,在幾副標(biāo)準(zhǔn)圖像增加高斯白噪聲圖像,一個(gè)非常簡單的復(fù)合脊波系數(shù)硬閾值的使用。

22、實(shí)驗(yàn)結(jié)果表明,復(fù)合脊波能夠提供比VisuShrink,Wiener 2和普通脊波更佳的去噪效果。我們建議ComRidgeletShrink用于實(shí)際的圖像去噪中。未來工作主要是考慮在復(fù)雜圖像應(yīng)用曲波復(fù)雜脊波。同樣,復(fù)雜脊波還可以應(yīng)用的不變特征提取模式識別方法。</p><p>  數(shù)字圖像處理方法的研究</p><p><b>  1 緒論</b></p>

23、<p>  數(shù)字圖像處理方法的研究源于兩個(gè)主要應(yīng)用領(lǐng)域:其一是為了便于人們分析而對圖像信息進(jìn)行改進(jìn);其二是為了使機(jī)器自動理解而對圖像數(shù)據(jù)進(jìn)行存儲、傳輸及顯示。</p><p>  1.1 數(shù)字圖像處理的概念</p><p>  一幅圖像可定義為一個(gè)二維函數(shù)f(x, y),這里x和y是空間坐標(biāo),而在任何一對空間坐標(biāo)f(x, y)上的幅值f稱為該點(diǎn)圖像的強(qiáng)度或灰度。當(dāng)x,y和幅值f

24、為有限的、離散的數(shù)值時(shí),稱該點(diǎn)是由有限的元素組成的,沒一個(gè)元素都有一個(gè)特定的位置和幅值,這些元素稱為圖像元素、畫面元素或象素。象素是廣泛用于表示數(shù)字圖像元素的詞匯。在第二章,將用更正式的術(shù)語研究這些定義。</p><p>  視覺是人類最高級的感知器官,所以,毫無疑問圖像在人類感知中扮演著最重要的角色。然而,人類感知只限于電磁波譜的視覺波段,成像機(jī)器則可覆蓋幾乎全部電磁波譜,從伽馬射線到無線電波。它們可以對非人類

25、習(xí)慣的那些圖像源進(jìn)行加工,這些圖像源包括超聲波、電子顯微鏡及計(jì)算機(jī)產(chǎn)生的圖像。因此,數(shù)字圖像處理涉及各種各樣的應(yīng)用領(lǐng)域。</p><p>  圖像處理涉及的范疇或其他相關(guān)領(lǐng)域(例如,圖像分析和計(jì)算機(jī)視覺)的界定在初創(chuàng)人之間并沒有一致的看法。有時(shí)用處理的輸人和輸出內(nèi)容都是圖像這一特點(diǎn)來界定圖像處理的范圍。我們認(rèn)為這一定義僅是人為界定和限制。例如,在這個(gè)定義下,甚至最普通的計(jì)算一幅圖像灰度平均值的工作都不能算做是圖像

26、處理。另一方面,有些領(lǐng)域(如計(jì)算機(jī)視覺)研究的最高目標(biāo)是用計(jì)算機(jī)去模擬人類視覺,包括理解和推理并根據(jù)視覺輸人采取行動等。這一領(lǐng)域本身是人工智能的分支,其目的是模仿人類智能。人工智能領(lǐng)域處在其發(fā)展過程中的初期階段,它的發(fā)展比預(yù)期的要慢得多,圖像分析(也稱為圖像理解)領(lǐng)域則處在圖像處理和計(jì)算機(jī)視覺兩個(gè)學(xué)科之間。</p><p>  從圖像處理到計(jì)算機(jī)視覺這個(gè)連續(xù)的統(tǒng)一體內(nèi)并沒有明確的界線。然而,在這個(gè)連續(xù)的統(tǒng)一體中可

27、以考慮三種典型的計(jì)算處理(即低級、中級和高級處理)來區(qū)分其中的各個(gè)學(xué)科。低級處理涉及初級操作,如降低噪聲的圖像預(yù)處理,對比度增強(qiáng)和圖像尖銳化。低級處理是以輸人、輸出都是圖像為特點(diǎn)的處理。中級處理涉及分割〔 把圖像分為不同區(qū)域或目標(biāo)物)以及縮減對目標(biāo)物的描述,以使其更適合計(jì)算機(jī)處理及對不同日標(biāo)的分類(識別)。中級圖像處理是以輸人為圖像,但輸出是從這些圖像中提取的特征(如邊緣、輪廓及不同物體的標(biāo)識等)為特點(diǎn)的。最后,高級處理涉及在圖像分析中

28、被識別物體的總體理解,以及執(zhí)行與視覺相關(guān)的識別函數(shù)(處在連續(xù)統(tǒng)一體邊緣)等。</p><p>  根據(jù)上述討論,我們看到,圖像處理和圖像分析兩個(gè)領(lǐng)域合乎邏輯的重疊區(qū)域是圖像中特定區(qū)域或物體的識別這一領(lǐng)域。這樣,在本書中,我們界定數(shù)字圖像處理包括輸人和輸出均是圖像的處理,同時(shí)也包括從圖像中提取特征及識別特定物體的處理。舉一個(gè)簡單的文本自動分析方面的例子來具體說明這一概念。在自動分析文本時(shí)首先獲取一幅包含文本的圖像,

29、對該圖像進(jìn)行預(yù)處理,提取(分割)字符,然后以適合計(jì)算機(jī)處理的形式描述這些字符,最后識別這些字符,而所有這些操作都在本書界定的數(shù)字圖像處理的范圍內(nèi)。理解一頁的內(nèi)容可能要根據(jù)理解的復(fù)雜度從圖像分析或計(jì)算機(jī)視覺領(lǐng)域考慮問題。這樣,本書定義的數(shù)字圖像處理的概念將在有特殊社會和經(jīng)濟(jì)價(jià)值的領(lǐng)域內(nèi)通用。在以下各章展開的概念是那些應(yīng)用領(lǐng)域所用方法的基礎(chǔ)。</p><p>  1.2數(shù)字圖像處理的起源</p><

30、;p>  數(shù)字圖像處理最早的應(yīng)用之一是在報(bào)紙業(yè),當(dāng)時(shí),圖像第一次通過海底電纜從倫敦傳往紐約。早在20世紀(jì)20年代曾引入Btutlane電纜圖片傳輸系統(tǒng),把橫跨大西洋傳送一幅圖片所需的時(shí)間從一個(gè)多星期減少到3個(gè)小時(shí)。為了用電纜傳輸圖片,首先要進(jìn)行編碼,然后在接收端用特殊的打印設(shè)備重構(gòu)該圖片。圖1.1就是用這種方法傳送并利用電報(bào)打印機(jī)通過字符模擬中間色調(diào)還原出來的圖像。</p><p>  這些早期數(shù)字圖像視覺

31、質(zhì)量的改進(jìn)工作,涉及到打印過程的選擇和亮度等級的分布等問題。用于得到圖1.1的打印方法到1921年底就被徹底淘汰了,轉(zhuǎn)而支持一種基于光學(xué)還原的技術(shù),該技術(shù)在電報(bào)接收端用穿孔紙帶打出圖片。圖1.2就是用這種方法得到的圖像,對比圖1.1,它在色調(diào)質(zhì)量和分辨率方面的改進(jìn)都很明顯。</p><p>  圖1.1 1421年由電報(bào)打印機(jī)采用特殊字 圖1.2 1922年在信號兩次穿越大西洋后,&l

32、t;/p><p>  符在編碼紙帶中產(chǎn)生的數(shù)字圖像 從穿孔紙帶得到的數(shù)字圖像,可以</p><p>  ( McFalsne) 看出某些差錯(cuò) ( McFalsne)</p><p>  早期的Bartlane系統(tǒng)可以用5個(gè)灰度等級對圖像編碼,到1929年已

33、增加到15個(gè)等級。圖1.3所示的這種典型類型的圖像就是用15級色調(diào)設(shè)備得到的。在這一時(shí)期,由于引入了一種用編碼圖像紙帶去調(diào)制光束而使底片感光的系統(tǒng),明顯地改善了復(fù)原過程。</p><p>  剛才引用的數(shù)字圖像的例子并沒有考慮數(shù)字圖像處理的結(jié)果,這主要是因?yàn)闆]有涉及到計(jì)算機(jī)。因此,數(shù)字圖像處理的歷史與數(shù)字計(jì)算機(jī)的發(fā)展密切相關(guān)。事實(shí)上,數(shù)字圖像要求非常大的存儲和計(jì)算能力,因此數(shù)字圖像處理領(lǐng)域的發(fā)展必須依靠數(shù)字計(jì)算機(jī)

34、及數(shù)據(jù)存儲、顯示和傳輸?shù)认嚓P(guān)技術(shù)的發(fā)展。</p><p>  計(jì)算機(jī)的概念可追溯到5000多年前中國算盤的發(fā)明。近兩個(gè)世紀(jì)以來的一些發(fā)展也奠定了計(jì)算機(jī)的基礎(chǔ)。然而,現(xiàn)代計(jì)算機(jī)的基礎(chǔ)還要回溯到20世紀(jì)40年代由約翰·馮·諾依曼提出的兩個(gè)重要概念:(l)保存程序和數(shù)據(jù)的存儲器;(2)條件分支。這兩個(gè)概念是中央處理單元(CPU)的基礎(chǔ)。今天,它是計(jì)算機(jī)的心臟。從馮·諾依曼開始,引發(fā)了一系列

35、重要技術(shù)進(jìn)步,使得計(jì)算機(jī)以強(qiáng)大的功能用于數(shù)字圖像處理領(lǐng)域。</p><p>  簡單說,這些進(jìn)步可歸納為如下幾點(diǎn):</p><p>  (1)1948年貝爾實(shí)驗(yàn)室發(fā)明了晶體三極管;</p><p>  (2)20世紀(jì)50年代到20世紀(jì)60年代高級編程語言(如COBOL和FORTRAN)的開發(fā);</p><p>  (3)1958年得州儀器公司

36、發(fā)明了集成電路(IC);</p><p>  (4)20世紀(jì)60年代早期操作系統(tǒng)的發(fā)展;</p><p>  (5)20世紀(jì)70年代Intel公司開發(fā)了微處理器(由中央處理單元、存儲器和輸入輸出控制組成的單一芯片);</p><p>  (6)1981年IBM公司推出了個(gè)人計(jì)算機(jī);</p><p>  (7)20世紀(jì)70年代出現(xiàn)的大規(guī)模集成電

37、路(LI)所引發(fā)的元件微小化革命,20世紀(jì)80年代出現(xiàn)了YLSI(超大規(guī)模集成電路),現(xiàn)在已出現(xiàn)了ULSI。</p><p>  圖1.3在1929年從倫敦到紐約用15級色調(diào)設(shè)備通過電纜</p><p>  傳送的Cenerale Pershing和Foch的未經(jīng)修飾的照片</p><p>  伴隨著這些技術(shù)進(jìn)步,大規(guī)模的存儲和顯示系統(tǒng)也隨之發(fā)展起來。這兩者均是數(shù)字

38、圖像處理的基礎(chǔ)。</p><p>  第一臺可以執(zhí)行有意義的圖像處理任務(wù)的大型計(jì)算機(jī)出現(xiàn)在20世紀(jì)60年代早期。數(shù)字圖像處理技術(shù)的誕生可追溯至這一時(shí)期這些機(jī)器的使用和空間項(xiàng)目的開發(fā),這兩大發(fā)展把人們的注意力集中到數(shù)字圖像處理的潛能上。利用計(jì)算機(jī)技術(shù)改善空間探測器發(fā)回的圖像的工作,始于1964年美國加利福尼亞的噴氣推進(jìn)實(shí)驗(yàn)室。當(dāng)時(shí)由“旅行者7號”衛(wèi)星傳送的月球圖像由一臺計(jì)算機(jī)進(jìn)行了處理,以校正航天器上電視攝像機(jī)中各

39、種類型的圖像畸變。圖1.4顯示了由“旅行者7號”于1954年7月31日上午(東部白天時(shí)間)9點(diǎn)09分在光線影響月球表面前約17分鐘時(shí)攝取的第一張?jiān)虑驁D像[痕跡(稱為網(wǎng)狀痕跡)用于幾何校正,在第5章將討論該間題],這也是美國航天器取得的第一幅月球圖像?!奥眯姓?號”傳送的圖像可作為改善的增強(qiáng)和復(fù)原圖像(例如來自“探索者”登月一飛行、“水手號”系列空間探淵器及阿波羅載人登月飛行的圖像)方法的基礎(chǔ)。</p><p> 

40、 進(jìn)行空間應(yīng)用的同時(shí),數(shù)字圖像處理技術(shù)在20世紀(jì)60年代末和20世紀(jì)70年代初開始用于醫(yī)學(xué)圖像、地球遙感監(jiān)測和天文學(xué)等領(lǐng)域。早在20世紀(jì)70年代發(fā)明的計(jì)算機(jī)軸向斷層術(shù)(CAT)[簡稱計(jì)算機(jī)斷層(CT)]是圖像處理在醫(yī)學(xué)診斷領(lǐng)域最重要的應(yīng)用之一。計(jì)算機(jī)軸向斷層術(shù)是一種處理方法,在這種處理中,一個(gè)檢測器環(huán)圍繞著一個(gè)物體(或病人),并且一個(gè)x射線源(與檢測器環(huán)同心)繞著物體旋轉(zhuǎn)。X射線穿過物體并由位于對面環(huán)中的相應(yīng)檢測器收集起來。當(dāng)X射線源旋

41、轉(zhuǎn)時(shí),重復(fù)這一過程。斷層技術(shù)由一些算法組成,該算法用感知的數(shù)據(jù)去重建通過物體的“切片”圖像。當(dāng)物體沿垂直于檢測器的方向運(yùn)動時(shí)就產(chǎn)生一系列這樣的“切片”,這些切片組成了物體內(nèi)部的再現(xiàn)圖像。斷層技術(shù)是由Godfrey N. Hounsfield先生和Allan M.Cormack教授發(fā)明的,他們共同獲得了1979年諾貝爾醫(yī)學(xué)獎(jiǎng)。X射線是在1895年由威廉·康拉德·倫琴發(fā)現(xiàn)的,由于這一發(fā)現(xiàn),他獲得了I901年諾貝爾物理學(xué)獎(jiǎng)

42、。這兩項(xiàng)發(fā)明相差近100年。它們在今天引領(lǐng)著圖像處理某些最活躍的應(yīng)用領(lǐng)域。</p><p>  圖1.4美國航天器傳送的第一張?jiān)虑蛘掌奥眯姓?號”</p><p>  衛(wèi)星1964年7月31日9點(diǎn)09分(東部白天時(shí)間)在</p><p>  光線影響月球表面前17分鐘時(shí)攝取的圖像</p><p>  Complex Ridgelets f

43、or Image Denoising</p><p>  G. Y. Chen and B. Kegl</p><p>  Abstract:The ridgelet transform based on the wavelet transform is a method of the multi-scale analysis. The digital composite ridgelet

44、 transform is adooted to remove the white noise embedded in the image,which is achieved by using a hard threshold of simple composite ridgelet coefficient.Experimental results show that the algorithm has better denoising

45、 effect than VisuShrink algorithm,ordinary ridgelet denoising algorithm and Wiener2 filter,which is provided buy the Matlab image processing toolb</p><p>  Keywords:image denoising; wavelet transform; ridgel

46、et transform; composite ridgelet </p><p>  1 Introduction</p><p>  Wavelet transforms have been successfully used in many scientific fields such as image compression, image denoising, signal pro

47、cessing, computer graphics,and pattern recognition, to name only a few.Donoho and his coworkers pioneered a wavelet denoising scheme by using soft thresholding and hard thresholding. This approach appears to be a good ch

48、oice for a number of applications. This is because a wavelet transform can compact the energy of the image to only a small number of large coefficients a</p><p>  2 Image Denoising by using Complex</p>

49、<p>  Ridgelets Discrete ridgelet transform provides near-ideal sparsity of representation of both smooth objects and of objects with edges. It is a near-optimal method of denoising for Gaussian noise. The ridgele

50、t transform can compress the energy of the image into a smaller number of ridgelet coe_cients. On the other hand, the wavelet transform produces many large wavelet coe_cients on the edges on every scale of the 2D wavelet

51、 decomposition. This means that many wavelet coe_cients are needed in ord</p><p>  1. Compute the 2D FFT of the image.</p><p>  2. Substitute the sampled values of the Fourier transform obtained

52、 on the square lattice with sampled values on a polar lattice.</p><p>  3. Compute the 1D inverse FFT on each angular line.</p><p>  4. Perform the 1D scalar wavelet transform on the resulting a

53、ngular lines in order to obtain the ridgelet coe_cients.</p><p>  It is well known that the ordinary discrete wavelet transform is not shift invariant because of the decimation operation during the transform

54、. A small shift in the input signal can cause very di_erent output wavelet coe_cients. In order to overcome this problem, Kingsbury introduced a new kind of wavelet transform, called the dual-tree complex wavelet transfo

55、rm, that exhibits approximate shift invariant property and improved angular resolution. Since the scalar wavelet is not shift invariant, it</p><p>  The complex ridgelet transform can be applied to the entir

56、e image or we can partition the image into a number of overlapping squares and we apply the ridgelet transform to each square. We decompose the original n _ n image into smoothly overlapping blocks of sidelength R pixels

57、 so that the overlap between two vertically adjacent blocks is a rectangular array of size R=2 _ R and the overlap between two horizontally adjacent blocks is a rectangular array of size R _ R=2 . For an n _ n image, we

58、co</p><p>  The thresholding for the complex ridgelet transform is similar to the curvelet thresholding [10]. One difference is that we take the magnitude of the complex ridgelet coe_cients when we do the th

59、resholding. Let y_ be the noisy ridgelet coe_cients. We use the following hard thresholding rule for estimating the unknown ridgelet coe_cients. When jy_j > k_~_, we let ^y_ = y_. Otherwise, ^y_ = 0. Here, ~It is appr

60、oximated by using Monte-Carlo simulations. The constant k used is dependent on the noise</p><p>  The complex ridgelet image denoising algorithm can be described as follows:</p><p>  1.Partition

61、 the image into R*R blocks with two vertically adjacent blocks overlapping R=2*R pixels and two horizontally adjacent blocks overlapping R _ R=2 pixels</p><p>  2. For each block, Apply the proposed complex

62、ridgelets, threshold the complex ridgelet coefficients, and perform inverse complex ridgelet transform.</p><p>  3. Take the average of the denoising image pixel values at the same location.</p><p

63、>  We call this algorithm ComRidgeletShrink,while the algorithm using the ordinary ridgelets RidgeletShrink. The computational complexity of ComRidgeletShrink is similar to that of RidgeletShrink by using the scalar w

64、avelets. The only di_erence is that we replaced the 1D wavelet transform with the 1D dual-tree complex wavelet transform. The amount of computation for the 1D dual-tree complex wavelet is twice that of the 1D scalar wave

65、let transform. However, other steps of the algorithm keep the same</p><p>  3 Experimental Results</p><p>  We perform our experiments on the well-known image Lena. We get this image from the fr

66、ee software package WaveLab developed by Donoho et al. at Stanford University. Noisy images with di_erent noise levels are generated by adding Gaussian white noise to the original noise-free images. For comparison, we im

67、plement VisuShrink, RidgeletShrink, ComRidgeletShrink and wiener2. VisuShrink is the universal soft-thresholding denoising technique. The wiener2 function is available in the MATLAB Image Proces</p><p>  4 C

68、onclusions and Future Work</p><p>  In this paper, we study image denoising by using complex ridgelets. Our complex ridgelet transform is obtained by performing 1D dual-tree complex wavelet onto the Radon tr

69、ansform coe_cients. The Radon transform is done by means of the projection-slice theorem. The approximate shift invariant property of the dual-tree complex wavelet transform makes the complex ridgelet transform an excell

70、ent choice for image denoising. The complex ridgelet transform provides near-ideal sparsity of representation</p><p>  The research of digital image processing technique </p><p>  1 Introductio

71、n</p><p>  Interest in digital image processing methods stems from two principal application areas: improvement of pictorial information for human interpretation; and processing of image data for storage, tr

72、ansmission, and representation for autonomous machine perception. This chapter has several objectives: (1)to define the scope of the field that we call image processing; (2)to give a historical perspective of the origins

73、 of this field; (3)to give an idea of the state of the art in image processing by ex</p><p>  1.1 What Is Digital Image Processing?</p><p>  An image may be defined as a two-dimensional functio

74、n, f(x, y), where x and y are spatial (plane) coordinates, and the amplitude of f at any pair of coordinates (x, y) is called the intensity or gray level of the image at that point. When x, y, and digital image. The fiel

75、d of digital image processing refers to processing digital images by means of a digital computer. Note that a digital image is composed of a finite number of elements, each of which has a particular location and value. T

76、hese ele</p><p>  Vision is the most advanced of our senses, so it is not surprising that images play the single most important role in human perception. However, unlike human who are limited to the visual b

77、and of the electromagnetic (EM) spectrum, imaging machines cover almost the entire EM spectrum, ranging from gamma to radio waves. They can operate on images generated by sources that human are not accustomed to associat

78、ing with image. These include ultrasound, electron microscopy, and computer-generated image</p><p>  There is no general agreement among authors regarding where image processing stops and other related areas

79、, such as image analysis and computer vision, start. Sometimes a distinction is made by defining image processing as a discipline in which both the input and output of a process are images. We believe this to be a limiti

80、ng and somewhat artificial boundary. For example, under this definition, even the trivial task of computing the average intensity of an image (which yields a single number) w</p><p>  There are no clear-cut

81、boundaries in the continuum from image processing at one end to computer vision at the other. However , one useful paradigm is to consider three types of computerized processes is this continuum: low-, mid-, and high-eve

82、r processes. Low-level processes involve primitive operation such as image preprocessing to reduce noise, contrast enhancement, and image sharpening. A low-level process is characterized by the fact that both its input a

83、nd output are images. Mid-level proce</p><p>  Based on the preceding comments, we see that a logical place of overlap between image processing and image analysis is the area of recognition of individual reg

84、ions or objects in an image. Thus, what we call in this book digital image processing encompasses processes whose inputs and outputs are images and, in addition, encompasses processes that extract attributes from images,

85、 up to and including the recognition of individual objects. As a simple illustration to clarify these concepts, consider</p><p>  1.2 The Origins of Digital Image Processing </p><p>  One of th

86、e first applications of digital images was in the newspaper industry, when pictures were first sent by submarine cable between London and NewYork. Introduction of the Bartlane cable picture transmission system in the ear

87、ly 1920s reduced the time required to transport a picture across the Atlantic from more than a week to less than three hours. Specialized printing equipment coded pictures for cable transmission and then reconstructed th

88、em at the receiving end. Figure 1.1 was transmitted</p><p>  Some of the initial problems in improving the visual quality of these early digital pictures were related to the selection of printing procedures

89、and the distribution of intensity levels. The printing method used to obtain Fig. 1.1 was abandoned toward the end of 1921 in favor of a technique based on photographic reproduction made from tapes perforated at the tele

90、graph receiving terminal. Figure 1.2 shows an images obtained using this method. The improvements over Fig. 1.1 are evident, both in t</p><p>  FIGURE 1.1 A digital picture produced in FIGURE 1.2 A digita

91、l picture</p><p>  1921 from a coded tape by a telegraph printer made in 1922 from a tape punched with special type faces (McFarlane) after the signals had crossed the Atlantic twice. Some errors are Visible

92、. (McFarlane)</p><p>  The early Bartlane systems were capable of coding images in five distinct level of gray. This capability was increased to 15 levels in 1929. Figure 1.3 is typical of the images that co

93、uld be obtained using the 15-tone equipment. During this period, introduction of a system for developing a film plate via light beams that were modulated by the coded picture tape improved the reproduction process consid

94、erably.</p><p>  Although the examples just cited involve digital images, they are not considered digital image processing results in the context of our definition because computer were not involved in their

95、 creation. Thus, the history of digital processing is intimately tied to the development of the digital computer. In fact digital images require so much storage and computational power that progress in the field of digit

96、al image processing has been dependent on the development of digital computers of supporti</p><p>  The idea of a computer goes back to the invention of the abacus in Asia Minor, more than 5000 years ago. Mo

97、re recently, there were developments in the past two centuries that are the foundation of what we call computer today. However, the basis for what we call a modern digital computer dates back to only the 1940s with the i

98、ntroduction by John von Neumann of two key concepts: (1) a memory to hold a stored program and data, and (2) conditional branching. There two ideas are the foundation of a c</p><p>  the invention of the tra

99、nsistor by Bell Laboratories in 1948;</p><p>  the development in the 1950s and 1960s of the high-level programming languages COBOL (Common Business-Oriented Language) and FORTRAN ( Formula Translator); <

100、/p><p>  the invention of the integrated circuit (IC) at Texas Instruments in 1958;</p><p>  the development of operating system in the early 1960s;</p><p>  the development of the mic

溫馨提示

  • 1. 本站所有資源如無特殊說明,都需要本地電腦安裝OFFICE2007和PDF閱讀器。圖紙軟件為CAD,CAXA,PROE,UG,SolidWorks等.壓縮文件請下載最新的WinRAR軟件解壓。
  • 2. 本站的文檔不包含任何第三方提供的附件圖紙等,如果需要附件,請聯(lián)系上傳者。文件的所有權(quán)益歸上傳用戶所有。
  • 3. 本站RAR壓縮包中若帶圖紙,網(wǎng)頁內(nèi)容里面會有圖紙預(yù)覽,若沒有圖紙預(yù)覽就沒有圖紙。
  • 4. 未經(jīng)權(quán)益所有人同意不得將文件中的內(nèi)容挪作商業(yè)或盈利用途。
  • 5. 眾賞文庫僅提供信息存儲空間,僅對用戶上傳內(nèi)容的表現(xiàn)方式做保護(hù)處理,對用戶上傳分享的文檔內(nèi)容本身不做任何修改或編輯,并不能對任何下載內(nèi)容負(fù)責(zé)。
  • 6. 下載文件中如有侵權(quán)或不適當(dāng)內(nèi)容,請與我們聯(lián)系,我們立即糾正。
  • 7. 本站不保證下載資源的準(zhǔn)確性、安全性和完整性, 同時(shí)也不承擔(dān)用戶因使用這些下載資源對自己和他人造成任何形式的傷害或損失。

評論

0/150

提交評論