基于稀疏貪婪搜索的人臉畫(huà)像合成
發(fā)布時(shí)間：2018-08-19 20:03
【摘要】：人臉畫(huà)像合成是通過(guò)機(jī)器學(xué)習(xí)對(duì)照片和畫(huà)像之間的復(fù)雜映射關(guān)系建模進(jìn)而利用該模型從照片合成畫(huà)像的過(guò)程。畫(huà)像合成對(duì)刑偵破案和數(shù)字娛樂(lè)具有重要的應(yīng)用價(jià)值。例如,當(dāng)案件發(fā)生后,由于環(huán)境或硬件條件的制約,警方無(wú)法獲取犯罪嫌疑人的影像資料。此時(shí),畫(huà)家根據(jù)受害者或者目擊證人的描述繪制的素描畫(huà)像就成為犯罪嫌疑人照片的最優(yōu)替代。將警方數(shù)據(jù)庫(kù)中的身份證照片轉(zhuǎn)換成對(duì)應(yīng)的素描畫(huà)像,再利用犯罪嫌疑人的素描畫(huà)進(jìn)行檢索,從而縮小或鎖定犯罪嫌疑人的身份。另外,隨著社交媒體的發(fā)展,很多年輕人都希望自己的用戶頭像富有個(gè)性,因此各種風(fēng)格的素描畫(huà)便成為他們熱衷的選擇之一。此外,人臉畫(huà)像合成還可作為其他計(jì)算機(jī)視覺(jué)任務(wù)的重要組成部分,比如人臉畫(huà)像老化等�，F(xiàn)有基于機(jī)器學(xué)習(xí)的人臉畫(huà)像合成方法可分成兩大類：模型驅(qū)動(dòng)和數(shù)據(jù)驅(qū)動(dòng)的方法。本論文致力于研究數(shù)據(jù)驅(qū)動(dòng)的方法,針對(duì)現(xiàn)有數(shù)據(jù)驅(qū)動(dòng)方法存在的一些不足,如對(duì)測(cè)試照片要求嚴(yán)格、需要依賴大量的訓(xùn)練樣本等,進(jìn)行方法的創(chuàng)新。本文的主要?jiǎng)?chuàng)新點(diǎn)可以概括為：1. 提出一種基于多照片-畫(huà)像對(duì)的人臉畫(huà)像合成方法�，F(xiàn)有的數(shù)據(jù)驅(qū)動(dòng)方法只考慮局部搜索策略,導(dǎo)致無(wú)法順利合成測(cè)試照片獨(dú)有的非人臉因素。此外,局部搜索要求測(cè)試照片和訓(xùn)練集中的圖像對(duì)齊,限制了測(cè)試照片的要求。針對(duì)上述問(wèn)題,提出一種基于多照片-畫(huà)像對(duì)的人臉畫(huà)像合成方法：第一步利用稀疏編碼算法將圖像塊像素特征變成稀疏表示特征,提高算法對(duì)干擾的魯棒性；第二步利用稀疏表示中每個(gè)稀疏系數(shù)的值以及稀疏系數(shù)編碼的次序這兩個(gè)信息對(duì)訓(xùn)練圖像塊建立搜索樹(shù),提高算法的搜索精度和速度；第三步利用測(cè)試照片的先驗(yàn)信息,結(jié)合圖模型,通過(guò)貝葉斯推斷進(jìn)行人臉畫(huà)像的合成。第一步和第二步簡(jiǎn)稱稀疏貪婪搜索算法。實(shí)驗(yàn)表明所提方法相較于已有的數(shù)據(jù)驅(qū)動(dòng)方法能更好更快的合成非人臉因素,而且對(duì)于任意的測(cè)試照片都適用。2. 提出一種基于單照片-畫(huà)像對(duì)的人臉畫(huà)像合成方法。現(xiàn)有數(shù)據(jù)驅(qū)動(dòng)方法需要依賴大量的照片-畫(huà)像對(duì)作為訓(xùn)練集。然而獲取大量照片-畫(huà)像對(duì)的代價(jià)很大,限制了已有方法的實(shí)際應(yīng)用。此外,在某些極端情況下只有一個(gè)照片-畫(huà)像對(duì)可用。針對(duì)上述問(wèn)題,提出一種基于單照片-畫(huà)像對(duì)的人臉畫(huà)像合成方法：第一步對(duì)訓(xùn)練集中的單照片-畫(huà)像對(duì)建立高斯金字塔,不僅增加了訓(xùn)練樣本而且考慮了人臉結(jié)構(gòu)的尺度信息；第二步利用稀疏貪婪搜索算法得到測(cè)試照片的初始畫(huà)像,充分保持了基于多照片-畫(huà)像對(duì)的人臉畫(huà)像合成方法的優(yōu)點(diǎn)；第三步利用由測(cè)試照片和初始畫(huà)像以及已有的單照片-畫(huà)像對(duì)所構(gòu)成的新訓(xùn)練集,通過(guò)結(jié)合級(jí)聯(lián)回歸策略和圖模型進(jìn)行最終的人臉畫(huà)像合成。實(shí)驗(yàn)表明所提方法能取得與最新數(shù)據(jù)驅(qū)動(dòng)方法可比擬的結(jié)果,而且同樣能合成非人臉因素且不限制測(cè)試照片的要求。3. 提出一種基于單目標(biāo)畫(huà)像的人臉畫(huà)像合成方法�，F(xiàn)有數(shù)據(jù)驅(qū)動(dòng)方法需要依賴照片-畫(huà)像對(duì),不管是大量還是一對(duì),都限制了合成任意風(fēng)格畫(huà)像的能力。針對(duì)上述問(wèn)題,提出一種基于單目標(biāo)畫(huà)像的人臉畫(huà)像合成方法。第一步利用稀疏貪婪搜索算法合成測(cè)試照片的初始畫(huà)像；第二步利用多尺度特征尋找候選畫(huà)像塊；第三步利用基于多特征的最優(yōu)化模型精選候選畫(huà)像塊；第四步利用級(jí)聯(lián)回歸策略對(duì)初始畫(huà)像進(jìn)行質(zhì)量提升。實(shí)驗(yàn)表明所提方法能取得與最新數(shù)據(jù)驅(qū)動(dòng)方法可比擬的結(jié)果。而且在以實(shí)驗(yàn)所列的風(fēng)格目標(biāo)畫(huà)像作為訓(xùn)練集的情況下,所提方法對(duì)于任意給定的測(cè)試照片都能合成質(zhì)量良好的對(duì)應(yīng)風(fēng)格畫(huà)像,這使算法更加有利于數(shù)字娛樂(lè)。4. 提出一種基于統(tǒng)一框架的人臉畫(huà)像合成方法�，F(xiàn)有數(shù)據(jù)驅(qū)動(dòng)方法在候選圖像塊搜索時(shí)只利用了局部搜索策略而上述所提方法則只利用了全局搜索策略。此外,已有方法在最終畫(huà)像合成時(shí)大多利用了多個(gè)候選塊的線性組合,導(dǎo)致合成結(jié)果有些平滑。而已有的高頻重建策略都是模型驅(qū)動(dòng)的方法,缺失畫(huà)像風(fēng)格信息。針對(duì)上述問(wèn)題,提出一種基于統(tǒng)一框架的人臉畫(huà)像合成方法：第一步將訓(xùn)練集劃分為初始訓(xùn)練集和高頻訓(xùn)練集；第二步給定測(cè)試照片利用局部搜索策略和全局搜索策略在初始訓(xùn)練集中進(jìn)行候選圖像塊的搜索,充分利用局部位置信息和人臉全局相似性的信息,然后利用圖模型進(jìn)行人臉初始畫(huà)像合成；第三步對(duì)給定的測(cè)試照片利用與初始畫(huà)像合成相同的策略進(jìn)行人臉高頻畫(huà)像的合成。將初始畫(huà)像和高頻畫(huà)像相加得到最終的人臉畫(huà)像。實(shí)驗(yàn)表明所提算法能同時(shí)合成非人臉部件和人臉部件,并且高頻細(xì)節(jié)更加豐富。綜上,本文本著訓(xùn)練樣本從多到少、研究思路從偏到全的方式,提出以稀疏貪婪搜索作為基礎(chǔ)的四種人臉畫(huà)像合成方法,用于提升人臉畫(huà)像合成的實(shí)用性。理論分析和實(shí)驗(yàn)結(jié)果表明了所提出方法相對(duì)于已有方法的優(yōu)越性。
[Abstract]:Face portrait synthesis is a process of synthesizing images from photographs by modeling the complex mapping relationship between photographs and images through machine learning. Portrait synthesis has important application value in criminal investigation and digital entertainment. For example, when a case occurs, the police can not obtain a crime because of environmental or hardware constraints. At this point, the painter's sketch based on the description of the victim or witness becomes the best alternative to the photograph of the criminal suspect. In addition, with the development of social media, many young people want their user portraits to be personalized, so sketching of various styles has become one of their favorite choices. In addition, face portrait synthesis can also be an important part of other computer vision tasks, such as face portrait aging. Machine-learning face portrait synthesis methods can be divided into two categories: model-driven and data-driven methods. This paper focuses on data-driven methods, aiming at the shortcomings of existing data-driven methods, such as strict requirements on test photos, relying on a large number of training samples, and so on, to innovate the methods. The main innovations can be summarized as follows: 1. A multi-photo-image pair based face image synthesis method is proposed. The existing data-driven methods only consider the local search strategy, which makes it impossible to successfully synthesize the non-face factors unique to the test photos. In addition, the local search requires the alignment of the test photos and the training set, which limits the test. In order to solve the above problems, this paper proposes a face image synthesis method based on multi-photo-image pairs. The first step is to use sparse coding algorithm to transform the pixel features of image blocks into sparse representation features to improve the robustness of the algorithm to interference. The second step is to use the value of each sparse coefficient in sparse representation and sparse coefficient coding. The order of the two information sets up a search tree for the training image blocks to improve the search accuracy and speed of the algorithm. The third step uses the prior information of the test photos and combines with the graph model to synthesize the face image by Bayesian inference. Data-driven methods can synthesize non-face factors better and faster, and can be applied to any test photograph. 2. A face image synthesis method based on single photo-image pairs is proposed. In addition, in some extreme cases there is only one photo-portrait pair available. To solve the above problem, a face portrait synthesis method based on single photo-portrait pair is proposed. The first step is to build a Gaussian pyramid for the single photo-portrait pair in the training set, which not only increases the training sample but also considers the human being. In the second step, the sparse greedy search algorithm is used to obtain the initial portrait of the test photos, which fully maintains the advantages of the multi-photo-image pair based face portrait synthesis method. In the third step, the new training set composed of the test photos, the initial portrait and the existing single-photo-image pairs is utilized, and the combination level is adopted. Experiments show that the proposed method can achieve comparable results with the latest data-driven methods, and can also synthesize non-face factors without restricting the requirements of the test photos. 3. A face image synthesis method based on single-object portrait is proposed. In order to solve the above problem, a method of face portrait synthesis based on single-object portrait is proposed. First, the initial portrait of the test picture is synthesized by sparse greedy search algorithm; second, the multi-scale feature is used to find the condition. In the third step, the candidate blocks are selected by the Multi-feature-based optimization model, and in the fourth step, the quality of the initial portrait is improved by cascade regression strategy. In the case of any given test image, the proposed method can synthesize good quality corresponding style portraits, which makes the algorithm more conducive to digital entertainment. 4. A unified framework based face portrait synthesis method is proposed. In addition, most of the existing methods utilize the linear combination of multiple candidate blocks in the final portrait synthesis, resulting in smoother results. However, the existing high-frequency reconstruction strategies are model-driven and lack of portrait style information. The first step is to divide the training set into the initial training set and the high frequency training set; the second step is to search the candidate image blocks in the initial training set by using local search strategy and global search strategy for a given test photo, making full use of the information of local location information and global face similarity, and then using the graph model to advance. In the third step, we use the same strategy as the original portrait synthesis to synthesize the high-frequency portrait of the human face for a given test picture. The final portrait is obtained by adding the initial portrait and the high-frequency portrait. In summary, this paper proposes four face image synthesis methods based on sparse greedy search to improve the practicability of face image synthesis. Theoretical analysis and experimental results show that the proposed method is superior to the existing methods.
【學(xué)位授予單位】：西安電子科技大學(xué)
【學(xué)位級(jí)別】：博士
【學(xué)位授予年份】：2016
【分類號(hào)】：TP391.41
本文編號(hào)：2192694
[Abstract]:Face portrait synthesis is a process of synthesizing images from photographs by modeling the complex mapping relationship between photographs and images through machine learning. Portrait synthesis has important application value in criminal investigation and digital entertainment. For example, when a case occurs, the police can not obtain a crime because of environmental or hardware constraints. At this point, the painter's sketch based on the description of the victim or witness becomes the best alternative to the photograph of the criminal suspect. In addition, with the development of social media, many young people want their user portraits to be personalized, so sketching of various styles has become one of their favorite choices. In addition, face portrait synthesis can also be an important part of other computer vision tasks, such as face portrait aging. Machine-learning face portrait synthesis methods can be divided into two categories: model-driven and data-driven methods. This paper focuses on data-driven methods, aiming at the shortcomings of existing data-driven methods, such as strict requirements on test photos, relying on a large number of training samples, and so on, to innovate the methods. The main innovations can be summarized as follows: 1. A multi-photo-image pair based face image synthesis method is proposed. The existing data-driven methods only consider the local search strategy, which makes it impossible to successfully synthesize the non-face factors unique to the test photos. In addition, the local search requires the alignment of the test photos and the training set, which limits the test. In order to solve the above problems, this paper proposes a face image synthesis method based on multi-photo-image pairs. The first step is to use sparse coding algorithm to transform the pixel features of image blocks into sparse representation features to improve the robustness of the algorithm to interference. The second step is to use the value of each sparse coefficient in sparse representation and sparse coefficient coding. The order of the two information sets up a search tree for the training image blocks to improve the search accuracy and speed of the algorithm. The third step uses the prior information of the test photos and combines with the graph model to synthesize the face image by Bayesian inference. Data-driven methods can synthesize non-face factors better and faster, and can be applied to any test photograph. 2. A face image synthesis method based on single photo-image pairs is proposed. In addition, in some extreme cases there is only one photo-portrait pair available. To solve the above problem, a face portrait synthesis method based on single photo-portrait pair is proposed. The first step is to build a Gaussian pyramid for the single photo-portrait pair in the training set, which not only increases the training sample but also considers the human being. In the second step, the sparse greedy search algorithm is used to obtain the initial portrait of the test photos, which fully maintains the advantages of the multi-photo-image pair based face portrait synthesis method. In the third step, the new training set composed of the test photos, the initial portrait and the existing single-photo-image pairs is utilized, and the combination level is adopted. Experiments show that the proposed method can achieve comparable results with the latest data-driven methods, and can also synthesize non-face factors without restricting the requirements of the test photos. 3. A face image synthesis method based on single-object portrait is proposed. In order to solve the above problem, a method of face portrait synthesis based on single-object portrait is proposed. First, the initial portrait of the test picture is synthesized by sparse greedy search algorithm; second, the multi-scale feature is used to find the condition. In the third step, the candidate blocks are selected by the Multi-feature-based optimization model, and in the fourth step, the quality of the initial portrait is improved by cascade regression strategy. In the case of any given test image, the proposed method can synthesize good quality corresponding style portraits, which makes the algorithm more conducive to digital entertainment. 4. A unified framework based face portrait synthesis method is proposed. In addition, most of the existing methods utilize the linear combination of multiple candidate blocks in the final portrait synthesis, resulting in smoother results. However, the existing high-frequency reconstruction strategies are model-driven and lack of portrait style information. The first step is to divide the training set into the initial training set and the high frequency training set; the second step is to search the candidate image blocks in the initial training set by using local search strategy and global search strategy for a given test photo, making full use of the information of local location information and global face similarity, and then using the graph model to advance. In the third step, we use the same strategy as the original portrait synthesis to synthesize the high-frequency portrait of the human face for a given test picture. The final portrait is obtained by adding the initial portrait and the high-frequency portrait. In summary, this paper proposes four face image synthesis methods based on sparse greedy search to improve the practicability of face image synthesis. Theoretical analysis and experimental results show that the proposed method is superior to the existing methods.
【學(xué)位授予單位】：西安電子科技大學(xué)
【學(xué)位級(jí)別】：博士
【學(xué)位授予年份】：2016
【分類號(hào)】：TP391.41
【相似文獻(xiàn)】
相關(guān)期刊論文 前10條
1 朱林;王士同;;人臉合成圖像檢測(cè)方法研究[J];計(jì)算機(jī)工程與應(yīng)用;2007年13期
2 張培;吳亞鋒;;基于局部指向性的反向合成圖像對(duì)齊算法[J];西北工業(yè)大學(xué)學(xué)報(bào);2007年04期
3 陳波;朱秋煜;;基于光照方向的人像合成圖像的檢測(cè)[J];計(jì)算機(jī)仿真;2009年03期
4 王偉;曾鳳;章國(guó)安;段新濤;;合成圖像被動(dòng)取證技術(shù)研究進(jìn)展[J];南通大學(xué)學(xué)報(bào)(自然科學(xué)版);2013年03期
5 錢霖,葉燕,臧崢寧;一種合成圖像的數(shù)字處理方法[J];蘇州大學(xué)學(xué)報(bào)(自然科學(xué));2002年02期
6 ;視角[J];科技傳播;2014年06期
7 劉元生;;通道合成圖像原理[J];影像技術(shù);2009年06期
8 侯志青;劉東州;康艷霜;劉立芳;唐志英;;Matlab6.0在拍合成中的輔助教學(xué)應(yīng)用[J];保定師范�？茖W(xué)校學(xué)報(bào);2006年04期
9 周維;汪增福;;用于高真實(shí)感口型合成的唇區(qū)肌肉模型[J];中國(guó)圖象圖形學(xué)報(bào);2008年11期
10 閻積惠;康慧;陳懷亮;田曉東;;甘肅北山TM合成圖像的波段組合模式與解譯[J];遙感信息;1989年01期
相關(guān)會(huì)議論文 前2條
1 曹奇志;;提高多光譜合成圖像質(zhì)量的一種新方法[A];中國(guó)光學(xué)學(xué)會(huì)2011年學(xué)術(shù)大會(huì)摘要集[C];2011年
2 毛慧蕓;龐家昊;金連文;杜明輝;;一種美麗人臉的計(jì)算機(jī)合成方法[A];第十五屆全國(guó)圖象圖形學(xué)學(xué)術(shù)會(huì)議論文集[C];2010年
相關(guān)博士學(xué)位論文 前1條
1 張聲傳;基于稀疏貪婪搜索的人臉畫(huà)像合成[D];西安電子科技大學(xué);2016年
相關(guān)碩士學(xué)位論文 前5條
1 張培;反向合成圖像對(duì)齊算法的研究及改進(jìn)[D];西北工業(yè)大學(xué);2007年
2 左龍;基于主動(dòng)表觀模型的多姿態(tài)人臉合成研究與應(yīng)用[D];西安電子科技大學(xué);2015年
3 郭超薇;數(shù)字接景真實(shí)感合成技術(shù)應(yīng)用研究[D];北京工業(yè)大學(xué);2013年
4 張嗣;造相[D];中央美術(shù)學(xué)院;2014年
5 田燦燦;半自動(dòng)感興趣區(qū)檢測(cè)算法在腎小球?yàn)V過(guò)率定量分析中的研究[D];上海交通大學(xué);2013年
，本文編號(hào)：2192694
本文鏈接：http://sikaile.net/shoufeilunwen/xxkjbs/2192694.html
最近更新
教材專著
