摘 要

紋理合成是圖像處理、計(jì)算機(jī)圖形學(xué)研究的重點(diǎn)領(lǐng)域之一，其目的是利用計(jì)算機(jī)通過算法產(chǎn)生高質(zhì)量的紋理。基于樣圖的紋理合成是最受關(guān)注的合成技術(shù)之一，它克服了傳統(tǒng)紋理合成過程中出現(xiàn)的種種缺點(diǎn)，利用小塊紋理樣本就能輸出大分辨率的紋理圖像，可廣泛應(yīng)用于計(jì)算機(jī)三維可視化、虛擬現(xiàn)實(shí)、高仿真游戲以及電影特技中。
提高紋理合成的速度和質(zhì)量一直是基于樣圖的紋理合成研究的兩個(gè)最基本的問題，近年來紋理合成的質(zhì)量有很大提高。因此，本文在保證紋理合成質(zhì)量的前提下，將如何提高紋理合成的效率和速度作為重點(diǎn)的研究?jī)?nèi)容。本文的主要工作和創(chuàng)新點(diǎn)如下：
1. 在研究基于MRF理論的紋理合成方法的基礎(chǔ)上，提出了一種基于SFLA-PSO混合算法的紋理合成方法，以加速紋理合成過程中最耗時(shí)的鄰域搜索匹配。實(shí)驗(yàn)結(jié)果表明，SFLA-PSO算法有PSO算法的簡(jiǎn)潔高效、搜索速度快的特點(diǎn)，同時(shí)又兼具SFLA全局搜索能力強(qiáng)，不易陷入局部最優(yōu)的優(yōu)點(diǎn)，應(yīng)用到紋理合成中效果顯著，性能穩(wěn)定。
2. 提出了一種塊自增的紋理合成方法，通過在新合成的紋理中選取尺寸不斷增加的紋理塊來分步加速合成圖像。由于紋理塊的尺寸不斷增加，合成的速度相較之固定尺寸的合成方法大幅提高，使得合成算法對(duì)于圖片分辨率的時(shí)間復(fù)雜度降低了一個(gè)函數(shù)階次。
3. 針對(duì)彩色紋理的合成，根據(jù)HVS理論進(jìn)一步改進(jìn)了灰度輔助合成的方法，使得算法更為方便簡(jiǎn)潔；對(duì)于紋理塊接縫處理，采取了權(quán)值融合的接縫平滑技術(shù)，相對(duì)于Efros的經(jīng)典縫合方法更加易于實(shí)現(xiàn)，計(jì)算量也相對(duì)較少；將上述各種加速手段結(jié)合起來，綜合運(yùn)用到紋理合成過程中，使得合成速度顯著提高。

關(guān)鍵詞 紋理合成；混合蛙跳；粒子群；塊拼接；權(quán)值融合；灰度輔助合成








































Abstract
Texture synthesis is one of the most important researching fields of image processing and computer graphics , what its purpose is to generate high quality texture by using computer through algorithm . Texture synthesising from samples is now the most concerned synthesis technology by the researchers, it overcame the shortcomings appeared in the process of traditional texture synthesis, Only using small texture samples can output large scale resolution texture image. This technology can be widely applied to computer 3D visualization, virtual reality, photo real video games and movie stunt, which makes it become one of the most popular directions of the researchers of computer graphic and image.
Improving the speed and quality of texture synthesis has been two basic issues of texture synthesis for years ,thinks for the efforts of the researchers , the quality of texture synthesis has been greatly improved. Therefore, this paper try to speed up the texture synthesis on the premise of guaranteed quality . The main work and innovations are as follows:
1.This paper analyzes the characteristics and limitations of MRF-based texture synthesis, present a SFLA-PSO algorithm of texture synthesis to accelerate the neighborhood search matching which is the most time-consuming process of texture synthesis. SFLA-PSO has the characteristic as PSO algorithm such like simple, efficient and fast, while also has the global search capability as SFLA ,not easy to fall into the local optimum in the neighborhood of texture synthesis, It performance well when applied to texture synthesis and works stabely.
2.This paper presents a block inreasing method to synthesis texture, the principle is that selecte increasing size of the block in the newly substep synthesized texture to synthesis finally image. Since the size of the texture blocks increases, the speed of synthesis is much faster compared with fixed size.
3.With color texture synthesis, we improved gray assisted synthesis algorithm further , making it more convenient and simple ;to deal with the blocks joint, this paper use weights blending technology to smooth the joint, as opposed to the classical Efros Image Quilting method it is more easy to implement, relatively few computing; this paper integrate the above methods to accelerate the synthesis process, making the synthesis rate increased significantly.

Keywords Texture Synthesis；Shuffled Frog Leaping Algorithm；Particle Swarm Optimization；Block Matching；Weight Blending；Gray Assistant Synthesis






























目 錄
摘 要 I
Abstract III
第1章 緒論 1
1.1 研究背景及意義 1
1.2 研究的歷史和現(xiàn)狀 2
1.3 論文的主要內(nèi)容和組織結(jié)構(gòu) 3
第2章 基于樣圖的紋理合成技術(shù) 5
2.1 引言 5
2.1.1 紋理的概念 5
2.1.2 紋理的分類 6
2.2 馬爾科夫隨機(jī)場(chǎng) 6
2.2.1 基于圖的MRF模型 8
2.2.2 基于圖的GRF模型 9
2.3 基于MRF的紋理合成算法 10
2.3.1 WL算法 10
2.3.2 Ashikhmin算法 12
2.4 基于塊拼接的紋理合成算法 13
2.4.1 Efros算法 14
2.4.2 LL算法 15
2.5 本章小結(jié) 17
第3章 基于SFLA-PSO的紋理合成算法 19
3.1 引言 19
3.2 PSO粒子群優(yōu)化算法 19
3.2.1 算法的數(shù)學(xué)描述 20
3.2.2 算法的實(shí)現(xiàn)和流程 22
3.2.3 算法的參數(shù)分析 23
3.3 基于PSO的紋理合成算法 25
3.3.1 算法原理 25
3.3.2 粒子的紋理意義 25
3.3.3 算法實(shí)現(xiàn) 26
3.3...