本文選題：圖像模糊去除 + 卷積神經(jīng)網(wǎng)絡(luò)�。� 參考：《安徽大學(xué)》2017年碩士論文

【摘要】：攝像機(jī)已經(jīng)滲透到人們生活的方方面面,圖像作為攝像機(jī)的產(chǎn)物,為人們傳遞信息提供了另一種重要途徑。然而由于種種原因,例如拍攝物體的運(yùn)動(dòng)、對(duì)焦不準(zhǔn)確、光照條件的不足等因素,造成拍攝得到的圖像是不清晰的。這些質(zhì)量退化的圖像往往不能滿足人類(lèi)的需求,嚴(yán)重時(shí)會(huì)造成一定的經(jīng)濟(jì)損失。去除模糊、復(fù)原圖像、提高圖像質(zhì)量成為人們研究工作關(guān)注的重點(diǎn)。本文主要針對(duì)去除圖像模糊與噪聲、恢復(fù)清晰圖像方面開(kāi)展研究。首先,介紹了基于卷積神經(jīng)網(wǎng)絡(luò)的圖像模糊去除研究的相關(guān)理論工作。分析圖像退化模型的基礎(chǔ)上,討論了幾種常用的求解模型,并詳細(xì)介紹了卷積神經(jīng)網(wǎng)絡(luò)的基本原理及面向圖像模糊去除的卷積神經(jīng)網(wǎng)絡(luò)模型。其次,依據(jù)前文的卷積神經(jīng)網(wǎng)絡(luò)的理論知識(shí),分析了現(xiàn)存的深度學(xué)習(xí)模糊去除方法的不足,詳細(xì)的介紹了本文設(shè)計(jì)的高頻信號(hào)保持且可快速模糊去除的快速卷積神經(jīng)網(wǎng)絡(luò)模型(fast CNN,FCNN)。在該網(wǎng)絡(luò)模型訓(xùn)練的過(guò)程中,對(duì)高頻圖像進(jìn)行傅里葉域上的梯度預(yù)處理,通過(guò)實(shí)施傅里葉域模糊去除的預(yù)處理得到一個(gè)初始的清晰圖像。接著將該初始圖像小塊作為輸入,相應(yīng)的真實(shí)清晰圖像小塊作為標(biāo)簽訓(xùn)練FCNN,得到從模糊圖像到潛在清晰圖像的映射函數(shù),實(shí)現(xiàn)基于該模型的模糊去除。預(yù)處理過(guò)程中,添加了梯度約束的高斯模型與提高平滑度的約束項(xiàng),將圖像的先驗(yàn)特征直接用于模糊去除,得到一個(gè)較魯棒的初值,為后續(xù)訓(xùn)練FCNN奠定基礎(chǔ)。FCNN模型由四層卷積層及激活函數(shù)構(gòu)成,這樣設(shè)計(jì)的目的在于移除圖像模糊的同時(shí),降低時(shí)間復(fù)雜度。實(shí)驗(yàn)結(jié)果表明,基于FCNN的圖像模糊去除方法相對(duì)于其他深度學(xué)習(xí)模糊去除方法,在有效的移除圖像模糊的基礎(chǔ)上,能夠更好的保持圖像的高頻紋理信息,同時(shí)降低了時(shí)間復(fù)雜度。最后,針對(duì)FCNN模型嚴(yán)重依賴(lài)梯度約束預(yù)處理導(dǎo)致FCNN方法自適應(yīng)較低的缺陷,提出增加模型深度方式得到改進(jìn)的FCNN模型—β-FCNN,并結(jié)合最小二乘方濾波預(yù)處理的方法實(shí)現(xiàn)一個(gè)新的去模糊方法,提高了自適應(yīng)性。首先考慮到卷積核大小為1×1的卷積層能夠在增加模型深度的情況下,最小化網(wǎng)絡(luò)訓(xùn)練參數(shù)的優(yōu)點(diǎn),對(duì)FCNN模型改進(jìn),得到β-FCNN模型;其次,考慮到最小二乘方濾波算法能夠使濾波后的圖像邊緣得到增強(qiáng)�；谠撎匦�,本文將最小二乘方濾波算法與β-FCNN模型結(jié)合(稱(chēng)之為β-FCNN方法)用于圖像模糊去除。網(wǎng)絡(luò)訓(xùn)練過(guò)程中,將最小二乘方濾波預(yù)處理后的圖像塊作為輸入,對(duì)應(yīng)的清晰圖像塊作為標(biāo)簽,訓(xùn)練該模型,得到從模糊圖像到潛在清晰圖像的映射函數(shù),實(shí)現(xiàn)基于該模型的圖像模糊去除。實(shí)驗(yàn)結(jié)果表明,β-FCNN方法相對(duì)于其他深度學(xué)習(xí)方法,有效的移除圖像模糊的基礎(chǔ)上,在一定范圍內(nèi),有較強(qiáng)的自適應(yīng)性。
[Abstract]:Cameras have penetrated into all aspects of people's lives. As a product of cameras, images provide another important way for people to transmit information. However, due to a variety of reasons, such as the motion of the shooting object, focusing inaccurate, insufficient lighting conditions and other factors, resulting in the image is not clear. These degraded images often can not meet the needs of human beings, and will cause certain economic losses. Removing blur, restoring images and improving image quality have become the focus of research. This paper focuses on removing image blur and noise and restoring clear image. Firstly, the theoretical work of image blur removal based on convolution neural network is introduced. Based on the analysis of image degradation model, several common solving models are discussed, and the basic principle of convolution neural network and the convolution neural network model for image fuzzy removal are introduced in detail. Secondly, according to the theoretical knowledge of convolution neural network, the shortcomings of the existing deep learning fuzzy removal methods are analyzed. The fast convolution neural network model designed in this paper is a fast fast convolution neural network model which can keep high frequency signal and can be removed quickly. In the process of training the network model, the gradient preprocessing of the high-frequency image is carried out in the Fourier domain, and an initial clear image is obtained by the pre-processing of the fuzzy removal in the Fourier domain. Then, the original image block is used as input and the corresponding real clear image block is used as label to train FCNN, and the mapping function from blurred image to latent clear image is obtained, and the fuzzy removal based on the model is realized. In the process of pretreatment, the Gao Si model with gradient constraint and the constraint item to improve the smoothness are added, and the prior features of the image are directly used to remove the blur, and a more robust initial value is obtained. The model is composed of four layers of convolution layer and activation function. The purpose of this design is to remove image blur and reduce the time complexity. The experimental results show that compared with other depth-learning fuzzy removal methods, the image fuzzy removal method based on FCNN can effectively remove the image blur, and can better keep the high-frequency texture information of the image. At the same time, the time complexity is reduced. Finally, in view of the disadvantage of FCNN model which depends heavily on gradient constraint preprocessing, the FCNN method is less adaptive. An improved FCNN model- 尾 -FCNN with increasing the depth of the model is proposed, and a new deblurring method is implemented by combining the least square filter preprocessing method, which improves the self-adaptability. Firstly, considering that the convolution layer with a convolution kernel size of 1 脳 1 can minimize the advantages of network training parameters while increasing the depth of the model, the 尾 -FCNN model is obtained by improving the FCNN model. Considering that the least square filter algorithm can enhance the edge of the filtered image. Based on this characteristic, the least square filtering algorithm and 尾 -FCNN model are combined in this paper for image blur removal. In the process of network training, the preprocessed image block of least square filter is taken as input, and the corresponding clear image block is used as label to train the model, and the mapping function from blurred image to latent clear image is obtained. Image blur removal based on this model is realized. The experimental results show that compared with other depth learning methods, the 尾 -FCNN method can effectively remove image blur and has strong adaptability in a certain range.
【學(xué)位授予單位】：安徽大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2017
【分類(lèi)號(hào)】：TP391.41;TP183

【相似文獻(xiàn)】

相關(guān)期刊論文 前10條

1 朱國(guó)超;趙素芹;;圖像模糊處理系統(tǒng)的設(shè)計(jì)與研究[J];軟件導(dǎo)刊;2008年03期

2 李強(qiáng);吳娟;;SHIMADZU WHA-200 C形臂圖像模糊調(diào)試[J];中國(guó)醫(yī)學(xué)裝備;2010年08期

3 張玲;掃描圖像模糊修復(fù)記[J];電腦愛(ài)好者;2002年04期

4 葛汝明;舊彩電圖像模糊故障的檢修[J];電子世界;1997年04期

5 王文穎;洪景新;顏益;;旋轉(zhuǎn)模糊圖像模糊尺度的確定方法[J];中國(guó)新通信;2009年03期

6 江曉;一種圖像模糊增強(qiáng)法[J];成都電訊工程學(xué)院學(xué)報(bào);1988年S2期

7 雷向康;一種改進(jìn)的圖像模糊增強(qiáng)方法[J];系統(tǒng)工程與電子技術(shù);1997年12期

8 李海芳;焦麗鵬;賀靜;;多特征綜合的圖像模糊情感注釋方法研究[J];中國(guó)圖象圖形學(xué)報(bào);2009年03期

9 日月光;;圖像模糊也玩花樣[J];電腦迷;2005年16期

10 張冰;;基于神經(jīng)網(wǎng)絡(luò)的圖像增強(qiáng)模型算法[J];計(jì)算機(jī)仿真;2011年12期

相關(guān)會(huì)議論文 前3條

1 徐彥明;李艦;段文義;尹健;;圖像模糊處理在雷達(dá)噪聲仿真實(shí)現(xiàn)中的應(yīng)用[A];第13屆中國(guó)系統(tǒng)仿真技術(shù)及其應(yīng)用學(xué)術(shù)年會(huì)論文集[C];2011年

2 施將君;劉進(jìn);劉軍;;質(zhì)子輻射照相的圖像模糊[A];第三屆全國(guó)加速器技術(shù)學(xué)術(shù)交流會(huì)論文摘要集[C];2007年

3 曹峰梅;陳喜春;金偉其;;基于極坐標(biāo)的相向運(yùn)動(dòng)圖像模糊的遞歸模型[A];光電技術(shù)與系統(tǒng)文選——中國(guó)光學(xué)學(xué)會(huì)光電技術(shù)專(zhuān)業(yè)委員會(huì)成立二十周年暨第十一屆全國(guó)光電技術(shù)與系統(tǒng)學(xué)術(shù)會(huì)議論文集[C];2005年

相關(guān)重要報(bào)紙文章 前6條

1 記者 楊圣泉;視頻圖像模糊也能揪出真兇[N];重慶商報(bào);2011年

2 記者 張星海;圖像模糊：關(guān)乎成敗[N];北京科技報(bào);2012年

3 上海 虞榮生;三洋彩電圖像模糊的檢修[N];電子報(bào);2006年

4 鼓浪嶼;超頻引起掃描圖像模糊[N];電腦報(bào);2004年

5 李寶春;長(zhǎng)虹精顯背投搶灘教育市場(chǎng)[N];中國(guó)消費(fèi)者報(bào);2002年

6 Yahgoro;Q與A[N];中國(guó)電腦教育報(bào);2003年

相關(guān)碩士學(xué)位論文 前4條

1 尹磊;圖像模糊復(fù)原及質(zhì)量評(píng)價(jià)算法研究[D];哈爾濱工業(yè)大學(xué);2015年

2 朱夢(mèng)宇;圖像模糊不變特征提取與識(shí)別技術(shù)研究[D];華中科技大學(xué);2015年

3 咸兆勇;圖像模糊檢測(cè)與模糊區(qū)域分割研究[D];廣西大學(xué);2013年

4 劉學(xué)平;低質(zhì)量圖像模糊人臉識(shí)別的研究[D];沈陽(yáng)航空工業(yè)學(xué)院;2010年

，

本文編號(hào)：1791705