基于数字蒙太奇的图像编辑方法研究

发布时间：2018-04-11 01:31

本文选题：数字蒙太奇 + 图分割　；参考：《浙江大学》2006年硕士论文

【摘要】：数字蒙太奇技术是指将同一场景的不同照片通过拼接、融合形成一幅新的更好地表达摄影师对场景理解的照片的一种架构技术。它作为一种新的图像处理技术和基于图像的绘制技术,近年来得到了广泛深入的研究,可广泛应用于科学研究、媒体制作、动漫设计等各个领域。作为一门新兴的技术,有必要对数字蒙太奇的技术及应用进行进一步的深入研究。本文详细介绍了有关基于数字蒙太奇图像编辑方法的研究和应用。在实现数字蒙太奇的图像编辑中,如何为合成照片的不同部分选择好的接缝,以及如何减少照片的可见拼接赝像,是两个关键问题。本文详细描述并采用了在计算机视觉领域广泛使用的图分割能量最小化理论,并利用最大流/最小割的算法计算能量最小的分割,以此为照片选择最佳的拼接映射。同时,本文使用了基于泊松(Poisson)方程的梯度域融合技术来消除或减少图像边缝拼接操作所遗留下来的可见赝像。本文的主要贡献在于,把蒙太奇图像编辑系统的应用划分为交互式和非交互式两种,说明了这两种情况的蒙太奇编辑系统在处理方法上的区别,并仔细介绍了图分割能量最小化理论和泊松图像编辑法在这两类不同情况下的具体实现细节。在上述理论和应用研究的基础上,我们提出了一种蒙太奇系统的设计框架和处理流程,并设计实现了一套基于数字蒙太奇的图像编辑系统。此系统实现了交互式和非交互式两大功能模块。非交互式的功能我们主要实现了扩展景深,运动物体的频闪可视化,净图版制作和背景重建等功能。在交互式的功能中,除了实现前人工作中已有的选择合成以外,我们还实现了较为完善的人像合成功能。此模块提供了创建人像库的功能,能够利用简单的用户交互,快速准确地对人像的脸部部位进行自动选择和边缘融合,达到较好的合成效果。用户也可以随时对人像库进行更新和检索。同时该系统提供了人性化的人机界面,简化了操作过程,具有良好的应用效果。
[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.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2006
【分类号】：TP391.41

【引证文献】