08-13, 15:30–16:10 (Asia/Yerevan), 213W PAB
Given a small example of a texture as an input the goal of texture synthesis is to generate an output image that is an expanded and smartly tiled version of the given input maintaining perceptual information. Texture synthesis methods are categorized into three main types: non-parametric, parametric and procedural methods. Two of these categories namely non-parametric and parametric are discussed during the talk. On the one hand, non-parametric approaches resample pixels or patches from the given source texture. Texture Optimization for Example-based Synthesis and Image Quilting for Texture Synthesis and Transfer are discussed. On the other hand, parametric methods require an explicit definition of a parametric texture. Two parametric methods, namely Texture Synthesis using CNNs and Non-Stationary Texture Synthesis by Adversarial Expansion are presented during the presentation. Results of the above-mentioned methods are demonstrated as a conclusion.
Given the structural definition of a texture as special variation in diverse layers of pixels demonstrating reiterating patterns combined with varied randomness in quantity the purpose of texture synthesis is to generate an expanded vision of the input texture that perceptually resembles the input. The goal of the presentation is to provide an overview of classical and neural texture synthesis algorithms. First, two classical non-parametric methods namely Texture Optimization for Example-based Synthesis and Image Quilting for Texture Synthesis and Transfer are covered. Second, two neural methods of texture synthesis are discussed: Texture Synthesis using CNNs and Non-Stationary Texture Synthesis by Adversarial Expansion. Third, advantages and disadvantages of these four methods are demonstrated. Results of the indicated four approaches and a visual comparison are provided.
Previous knowledge expected
Hovhannes Margaryan has a bachelor's degree in Computer Science from the American University of Armenia and is currently a master's student in Data Science at KU Leuven and ML Scientist at Picsart's Creative Intelligence team. Hovhannes is interested in different areas of computer vision and has experience with texture synthesis, style transfer, facial inpainting, image outpainting, full body generation, vector graphics, and physical simulation.