基于數(shù)字蒙太奇的圖像編輯方法研究
發(fā)布時(shí)間:2018-04-11 01:31
本文選題:數(shù)字蒙太奇 + 圖分割。 參考:《浙江大學(xué)》2006年碩士論文
【摘要】:數(shù)字蒙太奇技術(shù)是指將同一場(chǎng)景的不同照片通過(guò)拼接、融合形成一幅新的更好地表達(dá)攝影師對(duì)場(chǎng)景理解的照片的一種架構(gòu)技術(shù)。它作為一種新的圖像處理技術(shù)和基于圖像的繪制技術(shù),近年來(lái)得到了廣泛深入的研究,可廣泛應(yīng)用于科學(xué)研究、媒體制作、動(dòng)漫設(shè)計(jì)等各個(gè)領(lǐng)域。作為一門(mén)新興的技術(shù),有必要對(duì)數(shù)字蒙太奇的技術(shù)及應(yīng)用進(jìn)行進(jìn)一步的深入研究。 本文詳細(xì)介紹了有關(guān)基于數(shù)字蒙太奇圖像編輯方法的研究和應(yīng)用。在實(shí)現(xiàn)數(shù)字蒙太奇的圖像編輯中,如何為合成照片的不同部分選擇好的接縫,以及如何減少照片的可見(jiàn)拼接贗像,是兩個(gè)關(guān)鍵問(wèn)題。本文詳細(xì)描述并采用了在計(jì)算機(jī)視覺(jué)領(lǐng)域廣泛使用的圖分割能量最小化理論,并利用最大流/最小割的算法計(jì)算能量最小的分割,以此為照片選擇最佳的拼接映射。同時(shí),本文使用了基于泊松(Poisson)方程的梯度域融合技術(shù)來(lái)消除或減少圖像邊縫拼接操作所遺留下來(lái)的可見(jiàn)贗像。本文的主要貢獻(xiàn)在于,把蒙太奇圖像編輯系統(tǒng)的應(yīng)用劃分為交互式和非交互式兩種,說(shuō)明了這兩種情況的蒙太奇編輯系統(tǒng)在處理方法上的區(qū)別,并仔細(xì)介紹了圖分割能量最小化理論和泊松圖像編輯法在這兩類不同情況下的具體實(shí)現(xiàn)細(xì)節(jié)。 在上述理論和應(yīng)用研究的基礎(chǔ)上,我們提出了一種蒙太奇系統(tǒng)的設(shè)計(jì)框架和處理流程,并設(shè)計(jì)實(shí)現(xiàn)了一套基于數(shù)字蒙太奇的圖像編輯系統(tǒng)。此系統(tǒng)實(shí)現(xiàn)了交互式和非交互式兩大功能模塊。非交互式的功能我們主要實(shí)現(xiàn)了擴(kuò)展景深,運(yùn)動(dòng)物體的頻閃可視化,凈圖版制作和背景重建等功能。在交互式的功能中,除了實(shí)現(xiàn)前人工作中已有的選擇合成以外,我們還實(shí)現(xiàn)了較為完善的人像合成功能。此模塊提供了創(chuàng)建人像庫(kù)的功能,能夠利用簡(jiǎn)單的用戶交互,快速準(zhǔn)確地對(duì)人像的臉部部位進(jìn)行自動(dòng)選擇和邊緣融合,達(dá)到較好的合成效果。用戶也可以隨時(shí)對(duì)人像庫(kù)進(jìn)行更新和檢索。同時(shí)該系統(tǒng)提供了人性化的人機(jī)界面,簡(jiǎn)化了操作過(guò)程,具有良好的應(yīng)用效果。
[Abstract]:Digital montage technology is a kind of architecture technology, which combines different photos of the same scene to form a new and better representation of the photographer's understanding of the scene.As a new image processing technology and image-based rendering technology, it has been widely studied in recent years, and can be widely used in scientific research, media production, animation design and other fields.As a new technology, it is necessary to further study the technology and application of digital montage.This paper introduces the research and application of image editing method based on digital montage in detail.In the realization of digital montage image editing, two key problems are how to select good seams for different parts of composite photos and how to reduce visible stitching artifacts.This paper describes and adopts the minimization theory of image segmentation energy which is widely used in computer vision field, and uses the algorithm of maximum flow / minimum cut to calculate the segmentation of minimum energy, so as to select the best splicing map for photos.At the same time, the gradient domain fusion technique based on Poisson's Poisson equation is used to eliminate or reduce the visible artifacts left over by edge stitching.The main contribution of this paper is to divide the application of montage image editing system into interactive and non-interactive.The details of the theory of energy minimization of graph segmentation and Poisson's image editing method in these two different cases are introduced in detail.On the basis of the above theoretical and applied research, we propose a design framework and processing flow of montage system, and design and implement a digital montage based image editing system.This system realizes interactive and non-interactive function modules.The non-interactive functions include extended depth of field, stroboscopic visualization of moving objects, net plate making and background reconstruction.In the interactive function, in addition to the existing alternative synthesis in previous work, we also achieve a more complete function of portrait synthesis.This module provides the function of creating the human image library and can use simple user interaction to select and fuse the facial part of the image automatically and accurately, so as to achieve a better synthesis effect.Users can also update and retrieve the image library at any time.At the same time, the system provides a human-machine interface, simplifies the operation process, and has a good application effect.
【學(xué)位授予單位】:浙江大學(xué)
【學(xué)位級(jí)別】:碩士
【學(xué)位授予年份】:2006
【分類號(hào)】:TP391.41
【引證文獻(xiàn)】
相關(guān)期刊論文 前1條
1 張建橋;王長(zhǎng)元;;基于泊松方程的數(shù)字圖像無(wú)縫拼合[J];現(xiàn)代電子技術(shù);2010年17期
,本文編號(hào):1733858
本文鏈接:http://sikaile.net/wenyilunwen/dongmansheji/1733858.html
最近更新
教材專著