图像拼接：纹理合成与转移的新方法

4星 · 超过85%的资源需积分: 33 174 浏览量更新于2024-09-13 收藏 8.59MB PDF 举报

"图像拼接（Image Quilting）是一种创新的图像合成方法，它通过将现有图片的小块拼接起来创造出新的视觉效果。这项技术由Alexei A. Efros和William T. Freeman在论文中提出，他们分别来自加州大学伯克利分校和三菱电机研究实验室。论文的标题强调了其在纹理合成和转移方面的应用。首先，图像拼接作为一款快速且简单的纹理合成算法，展现出惊人的性能。它能够在广泛的纹理类型上产生高质量的结果，这在传统纹理合成方法之外开辟了一个新的可能性。通过这种方式，图像中的细节和纹理能够被有效地复现和融合，从而创造出独特的视觉体验。其次，该算法进一步扩展到纹理转移，即能够在不同对象间转移纹理。例如，研究人员可以将一种物体的纹理应用到另一个物体上，实现逼真的外观转换，这种技术在游戏开发、艺术创作和视觉特效等领域具有广泛应用。这种方法的优势在于它不需要3D模型，而是直接处理二维图像，降低了对三维数据的依赖。更广泛地讲，图像拼接展示了如何让一个图像在风格上模仿另一张图像，这在艺术模仿、风格迁移以及图像修复中都是一大突破。这种技术能够捕捉并保留源图像的特征，同时赋予目标图像以新的视觉语境或氛围。关键词：纹理合成、纹理映射、基于图像的渲染。这些关键词揭示了图像拼接的核心技术原理和相关领域。图像拼接是一种强大的工具，它不仅革新了纹理处理的方式，还推动了计算机图形学领域在图像处理和生成上的创新。"

Image Quilting for Texture Synthesis and Transfer

Alexei A. Efros

1,2

William T. Freeman

University of California, Berkeley

Mitsubishi Electric Research Laboratories

Abstract

We present a simple image-based method of generating novel vi-

sual appearance in which a new image is synthesized by stitching

together small patches of existing images. We call this process im-

age quilting. First, we use quilting as a fast and very simple texture

synthesis algorithm which produces surprisingly good results for

a wide range of textures. Second, we extend the algorithm to per-

form texture transfer – rendering an object with a texture taken from

a different object. More generally, we demonstrate how an image

can be re-rendered in the style of a different image. The method

works directly on the images and does not require 3D information.

Keywords: Texture Synthesis, Texture Mapping, Image-based

Rendering

1 Introduction

In the past decade computer graphics experienced a wave of ac-

tivity in the area of image-based rendering as researchers explored

the idea of capturing samples of the real world as images and us-

ing them to synthesize novel views rather than recreating the entire

physical world from scratch. This, in turn, fueled interest in image-

based texture synthesis algorithms. Such an algorithm should be

able to take a sample of texture and generate an unlimited amount

of image data which, while not exactly like the original, will be per-

ceived by humans to be the same texture. Furthermore, it would be

useful to be able to transfer texture from one object to anther (e.g.

the ability to cut and paste material properties on arbitrary objects).

In this paper we present an extremely simple algorithm to ad-

dress the texture synthesis problem. The main idea is to synthesize

new texture by taking patches of existing texture and stitching them

together in a consistent way. We then present a simple generaliza-

tion of the method that can be used for texture transfer.

1.1 Previous Work

Texture analysis and synthesis has had a long history in psychol-

ogy, statistics and computer vision. In 1950 Gibson pointed out

the importance of texture for visual perception [8], but it was the

pioneering work of Bela Julesz on texture discrimination [12] that

paved the way for the development of the ﬁeld. Julesz suggested

Computer Science Division, UC Berkeley, Berkeley, CA 94720, USA.

MERL, 201 Broadway, Cambridge, MA 02139, USA.

input images quilting results

Figure 1: Demonstration of quilting for texture synthesis and tex-

ture transfer. Using the rice texture image (upper left), we can syn-

thesize more such texture (upper right). We can also transfer the

rice texture onto another image (lower left) for a strikingly differ-

ent result.

that two texture images will be perceived by human observers to

be the same if some appropriate statistics of these images match.

This suggests that the two main tasks in statistical texture synthe-

sis are (1) picking the right set of statistics to match, (2) ﬁnding an

algorithm that matches them.

Motivated by psychophysical and computational models of hu-

man texture discrimination [2, 14], Heeger and Bergen [10] pro-

posed to analyze texture in terms of histograms of ﬁlter responses

at multiple scales and orientations. Matching these histograms it-

eratively was sufﬁcient to produce impressive synthesis results for

stochastic textures (see [22] for a theoretical justiﬁcation). How-

ever, since the histograms measure marginal, not joint, statistics

they do not capture important relationships across scales and ori-

entations, thus the algorithm fails for more structured textures. By

also matching these pairwise statistics, Portilla and Simoncelli [17]

were able to substantially improve synthesis results for structured

textures at the cost of a more complicated optimization procedure.

In the above approaches, texture is synthesized by taking a ran-

dom noise image and coercing it to have the same relevant statistics

as in the input image. An opposite approach is to start with an in-

put image and randomize it in such a way that only the statistics

to be matched are preserved. De Bonet [3] scrambles the input in

a coarse-to-ﬁne fashion, preserving the conditional distribution of

ﬁlter outputs over multiple scales (jets). Xu el.al. [21], inspired by

the Clone Tool in P

HOTOSHOP, propose a much simpler approach

yielding similar or better results. The idea is to take random square

blocks from the input texture and place them randomly onto the

synthesized texture (with alpha blending to avoid edge artifacts).

下载后可阅读完整内容，剩余5页未读，立即下载

w201026811412

粉丝: 0
资源: 2

图像拼接：纹理合成与转移的新方法

机器视觉基础 | TextureTransfer | 人脸纹理传输

纹理图像分割，包括一个pdf文件，几个m文件

纹理图像分割Matlab源代码 PDF PPT

Texture_Synthesis_Using_Image_Quilting.rar_image synthesis _quil

Texture.zip_Image Quilting _quilting matlab_texture_图像纹理_纹理合成

Algorithm-TextureSynthesis.zip

image quilting.rar

Image Quilting算法初学者

matlab 实现image quilt算法

quilting:用于创建，共享和修改被子设计的Web应用程序。 在服务器上的Node.js和MongoDB，客户端上的AngularJS和Socket.IO上运行以进行通信

最新资源

quilting:用于创建，共享和修改被子设计的Web应用程序。在服务器上的Node.js和MongoDB，客户端上的AngularJS和Socket.IO上运行以进行通信