Image and Color Stylization through Region-based Abstraction

It appears your Web browser is not configured to display PDF files. Download adobe Acrobat or click here to download the PDF file.

Click here to download the PDF file.

Creator: 

Azami, Rosa

Date: 

2021

Abstract: 

In Non-Photorealistic Rendering (NPR), abstraction methods simplify images. The NPR techniques produce stylized imagery by emphasizing the main structures through stroke elements, colors and removing unessential image information. However, the difficulty of abstraction lies in deciding what is important to be presented. This thesis introduces a few non-photorealistic rendering techniques for stylizing images by manipulating the image's elements such as details, complex region forms, and colors. To increase the visual interest in abstractions, it integrates contrast, texture, and structure preservation in designing the proposed algorithms. We propose an abstraction method based on an over-segmentation through overlapping region growth. The main idea of this method is to create stylized images by complex primitives. Unlike traditional over-segmentation methods, this method emphasizes the irregularity of region shape even in uniform areas. The region creation by overlapping allows us to preserve small structures. Additionally, we extend the creation of stylized images through this framework by manipulating the color and shapes of the region primitives. As an application of the region-based framework, we present a novel recoloring algorithm based on a tree-based structure, which produces wild recolorizations. A set of colorful abstractions are produced with various styles by considering contrast and structure preservation. We further engage image regions in providing a method for improving stippling in extreme-tone regions of the image. This method identifies dark regions and replaces them with polygons rather than stipples. For light regions, it suggests a mechanism to control the stipple placement. This approach generates stippled images with better visual quality than the originals despite using fewer stipples. Finally, we present an automatic system for photo manipulation that brightens an abstraction and alters the detail levels. By employing a smooth map, we brighten the photograph such that the areas of interest will get increased contrast while colors in the areas of less interest will become lighter and less saturated. We define a system for scoring texture elements, to establish different levels of texture indication.

Subject: 

Computer Science
Information Science

Language: 

English

Publisher: 

Carleton University

Thesis Degree Name: 

Doctor of Philosophy: 
Ph.D.

Thesis Degree Level: 

Doctoral

Thesis Degree Discipline: 

Computer Science

Parent Collection: 

Theses and Dissertations

Items in CURVE are protected by copyright, with all rights reserved, unless otherwise indicated. They are made available with permission from the author(s).