A genetic algorithm for texture synthesis and transfer

Other Title(s)

الخوارزمية الجينية في تركيب و نقل الصور النسيجية

Dissertant

Ahmad, Laylan Muhammad Rashid

Thesis advisor

Atiyyah, Bara Ali

Comitee Members

al-Abbudi, Bushra Qasim
Stephan, Jane Jalil

University

University of Baghdad

Faculty

College of Science

Department

Department of Computer Science

University Country

Iraq

Degree

Master

Degree Date

2005

English Abstract

Textures can describe a wide variety of natural phenomena with random variations over repeating patterns.

Examples of textures include images, motions, and surface geometry.

Textures are important for reproducing and rendering detailed surface appearance which is a major goal for computer graphics to achieve realism of the physical world.

A texture synthesis method starts from a sample image and attempts to produce a texture with visual appearance similar to that sample.

On the other hand, texture transfer takes two images -the source texture and the target controlling image- as input, and modifies the target image, replacing some high-frequency information with the source texture.

In this thesis, a new method for texture synthesis and transfer using genetic algorithm is proposed.

The computer technique of the genetic algorithm is applied to synthesize a new texture -represented in genotype level as genes of patches- given some source texture.

The algorithm can also be used to transfer patches of a source texture onto patches of a target controlling image.

The proposed genetic algorithm can be applied to render an artistic style or material object onto another image and can also be applied to enhance (increase resolution) of some images.

The proposed GA alters the mechanism of the main perturbation operator -crossover- so as to tackle texture synthesis and transfer problems.

The proposed Semi Uniform Crossover (SUX) turns the mechanism of the traditional uniform crossover into urgently spreading out vital genes (good patches) between offspring individuals at the expense of lethal or mortal genes rather than exchanging parental genes with fair probability.

Moreover, the mating process of the genetic algorithm is provided with two fashions: a bisexual recombination scheme where two parents are recombined at a single recombination step via local SUX operator and a multi-sexual recombination where a global SUX is applied among all individuals to form a new single offspring.

All other offspring are then obtained via mutation only.

Results of the proposed genetic algorithm are promising and quite comparable (in term of quality, efficiency, generality, simplicity and userfriendly) to other texture synthesis and transfer algorithms found in the literature including the current state-of-the-art algorithms of Efros and Freeman, Hertzmann et al, and Ashikhmin.

Main Subjects

Information Technology and Computer Science

Topics

• Algorithms
• Data processing

No. of Pages

111

Table of Contents

• APA
• MLA
• AMA

Language

English

Data Type

Arab Theses

Record ID

BIM-605583