本文關(guān)鍵詞： 動(dòng)漫圖像 圖像分割 語義級(jí)紋理推薦 紋理合成 光照定制　出處：《深圳大學(xué)》2017年碩士論文　論文類型：學(xué)位論文

【摘要】：伴隨網(wǎng)絡(luò)的普及以及動(dòng)漫產(chǎn)業(yè)自身的高速成長(zhǎng),動(dòng)漫的市場(chǎng)需求量也相應(yīng)的加大。然而動(dòng)漫的制作是一個(gè)繁瑣復(fù)雜的過程,動(dòng)漫產(chǎn)品的誕生需要長(zhǎng)時(shí)間的制作周期,這就導(dǎo)致了動(dòng)漫產(chǎn)業(yè)的高投入低產(chǎn)出。因此如何提高動(dòng)漫產(chǎn)品的產(chǎn)量,關(guān)鍵在于提高動(dòng)漫的制作技術(shù),縮短制作周期。經(jīng)過對(duì)動(dòng)漫制作的深入觀察和研究發(fā)現(xiàn),雖然每一幀漫畫都是人手畫出的,每一個(gè)細(xì)節(jié)都需要精心雕刻,但是,動(dòng)漫圖像還是存在一些重復(fù)出現(xiàn)的元素,比如沙灘、草坪、織物表面、墻紙、地表、樹的表皮等。這些區(qū)域的繪制是否可以用其他高效的方法替代?針對(duì)這一系列問題,本文提出動(dòng)漫圖像交互式紋理替換的方案,可以在很大程度上節(jié)約時(shí)間,提高效率。本文圍繞這個(gè)方案分別對(duì)以下三個(gè)方面展開研究。首先是圖像分割,進(jìn)行紋理替換的前提是將該區(qū)域與其他區(qū)域區(qū)分開,這就涉及圖像處理中的圖像分割。為此在對(duì)多種主流的圖像分割算法的深入分析研究之后,本文借助SLIC超像素的算法理念提出了全新的分割算法。在SLIC超像素生成算法的原理上,對(duì)該算法進(jìn)行了改進(jìn)和創(chuàng)新,得到更好的實(shí)驗(yàn)結(jié)果,使得分割效果更契合人類對(duì)分割的定義。其次是紋理推薦,在成千上萬的圖片中尋找合適的紋理本是一件繁瑣乏味的過程,紋理推薦可以很好的解決這個(gè)問題。本文在原有的紋理特征提取算法上,總結(jié)出一系列新的紋理度量值,例如紋理性、同質(zhì)性、重復(fù)性等。此外,我們還建立一個(gè)龐大的紋理樣圖庫。并結(jié)合這些紋理特征,實(shí)現(xiàn)語義級(jí)紋理推薦,通過設(shè)置紋理特征值,系統(tǒng)就可以檢索紋理圖庫推薦出相應(yīng)的紋理樣圖。最后是紋理合成,將紋理樣圖完美的替換到原圖上是一個(gè)關(guān)鍵的步驟,在綜合研究分析了多種紋理合成算法的各方面性能,本文選擇了Graphcut紋理合成算法,并對(duì)該算法實(shí)現(xiàn)并行化改進(jìn),并且在原算法的基礎(chǔ)上提出了光照定制這一創(chuàng)新點(diǎn),可以根據(jù)輸入光照?qǐng)D像來影響合成效果。在時(shí)間性和合成效果上,使得Graphcut算法都得到改進(jìn)。
[Abstract]:With the popularity of the network and the rapid growth of the animation industry itself, the market demand for animation is also increased accordingly. However, the production of animation is a complicated process. The birth of animation products needs a long production cycle, which leads to the animation industry high input and low output. So how to improve the production of animation products, the key is to improve the animation production technology. Shorten the production cycle. Through the animation production in-depth observation and research found that although each frame of cartoon is hand-drawn, every detail needs to be carefully carved, but. There are still some recurring elements in anime images, such as beach, lawn, fabric surface, wallpaper, surface, tree skin and so on. Can these areas be replaced by other efficient methods? Aiming at this series of problems, this paper proposes an interactive texture replacement scheme for animation images, which can save time to a great extent. This paper focuses on the following three aspects: firstly, image segmentation, the premise of texture replacement is to distinguish the region from other regions. This involves image segmentation in image processing. In this paper, a new segmentation algorithm is proposed based on the idea of SLIC super-pixel algorithm. In the principle of SLIC super-pixel generation algorithm, the algorithm is improved and innovated, and better experimental results are obtained. Make the segmentation more in line with the human definition of segmentation. Secondly, texture recommendation, looking for the appropriate texture in thousands of images is a tedious process. Texture recommendation can solve this problem very well. In this paper, a series of new texture measures, such as texture, homogeneity, repeatability and so on, are summarized in the original texture feature extraction algorithm. We also build a huge texture map library, and combine these texture features to realize the semantic level texture recommendation, by setting the texture feature value. The system can retrieve the texture database and recommend the corresponding texture image. Finally, the texture synthesis is the key step to replace the texture image to the original image perfectly. In the comprehensive research and analysis of the various aspects of the performance of texture synthesis algorithms, this paper selected the Graphcut texture synthesis algorithm, and the algorithm to achieve parallelization improvement. On the basis of the original algorithm, the innovation of illumination customization is put forward, which can affect the synthesis effect according to the input illumination image. In terms of timing and composition effect, the Graphcut algorithm is improved.
【學(xué)位授予單位】：深圳大學(xué)
【學(xué)位級(jí)別】：碩士
【學(xué)位授予年份】：2017
【分類號(hào)】：J218.2;TP391.41

【相似文獻(xiàn)】

相關(guān)會(huì)議論文 前8條

1 徐國(guó)棟;郭恒業(yè);;統(tǒng)計(jì)紋理生成模型[A];中國(guó)科學(xué)技術(shù)協(xié)會(huì)首屆青年學(xué)術(shù)年會(huì)論文集（工科分冊(cè)·上冊(cè)）[C];1992年

2 盧海燕;陳明;孫梅;宋樹良;馬艷;鄭傳立;;肝纖維化超聲圖像紋理分析與組織學(xué)檢查對(duì)照研究[A];第二屆長(zhǎng)三角超聲醫(yī)學(xué)論壇暨2009年浙江省超聲醫(yī)學(xué)學(xué)術(shù)年會(huì)論文匯編[C];2009年

3 楊關(guān);馮國(guó)燦;鄒小林;劉志勇;;彩色紋理分析中的高斯圖模型[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

4 熊欣;何中市;;紋理分析在超聲圖像中的應(yīng)用[A];2008年計(jì)算機(jī)應(yīng)用技術(shù)交流會(huì)論文集[C];2008年

5 姚明海;應(yīng)宏微;;基于反饋的汽車牌照紋理分析定位算法的研究[A];第二十三屆中國(guó)控制會(huì)議論文集（下冊(cè)）[C];2004年

6 黃戰(zhàn)華;劉正;張振;蔡懷宇;楊基春;張尹馨;;基于矢量化方法的紋理缺陷快速檢測(cè)[A];2008中國(guó)儀器儀表與測(cè)控技術(shù)進(jìn)展大會(huì)論文集（Ⅰ）[C];2008年

7 管玉娟;王淑晴;關(guān)澤群;;紋理分析的條件模式譜方法[A];第十四屆全國(guó)遙感技術(shù)學(xué)術(shù)交流會(huì)論文摘要集[C];2003年

8 劉佳;李哲英;王淑英;;路標(biāo)識(shí)別算法研究[A];2008通信理論與技術(shù)新發(fā)展——第十三屆全國(guó)青年通信學(xué)術(shù)會(huì)議論文集（下）[C];2008年

相關(guān)博士學(xué)位論文 前3條

1 高妮;基于Contourlet提取紋理建立早期AD的高斯過程預(yù)測(cè)模型[D];首都醫(yī)科大學(xué);2016年

2 毛羽;血清miR-1290聯(lián)合DW-MRI紋理分析預(yù)測(cè)食管鱗癌放化療敏感性及相關(guān)機(jī)制的研究[D];天津醫(yī)科大學(xué);2016年

3 張國(guó)鵬;三維紋理分析在虛擬結(jié)腸鏡計(jì)算機(jī)輔助檢測(cè)/診斷中的應(yīng)用研究[D];第四軍醫(yī)大學(xué);2013年

相關(guān)碩士學(xué)位論文 前10條

1 陳金霞;快速粒度估計(jì)法在紋理分析中的應(yīng)用研究[D];華南理工大學(xué);2015年

2 劉W，

本文編號(hào)：1479069