1. Basic information

Thesis title: LazyBrush: Painting Tool for Flexible Hand-drawn Cartoons

LazyBrush: a flexible coloring tool for hand-drawn cartoon color (compatible: comics, line art, drawing, animation)

Papers Link: (1) CVUT (2) Wiley Online Library

Research fields: computer graphics, based on the color line art coloring tips, interactive image segmentation (Interactive Image Segmentation), graph theory (Graph Theory) in "Figure cut problem" (Graph Cut)

作者：Daniel Sýkora, John Dingliana, Steven Collins - Trinity College Dublin

Published meeting: Eurographics 2009 (CCF Class B meeting)

Citations: 81 (as of 2020.03)

Code and Model:

(1) Official Code: only commercial closed-source implementation, see the project home page

(2) unofficial reproduction:

A. kosua20 / LazyBrush-Implementation , Matlab achieve high quality teaching demonstration projects, including French documentation

B. furaga / LazyBrushSharp , C # achieve

c. Evarin/LazyBrush

Significance:

1. The algorithms have been integrated into commercial painting software TVPaint, as well as open-source software in Krita.

2. Based on the traditional segmentation methods , which bring many advantages:

(1) robustness (generalization), the success rate is much higher than the color depth of learning ;

(2) small amount of calculation, color depth faster than the neural network;

(3) low hardware requirements: suitable for running on a single CPU, large memory need not have server-level GPU acceleration;

(4) Segmentation based on graph theory, without training data set, the training process need not.

3. generalization algorithm is very strong (Flexible), compatible with various types of hand-drawn cartoons color, including: cartoons, line art pencils, pens, line art, drawing, coloring and rapid continuous animation frames.

2. Overview of the whole idea of paper

Overall thesis is divided into two parts: the first part is the algorithm described demand, put forward the ideal of the perfect color characteristics of the algorithm to be met. The second part is the problem modeling and algorithm design.

The problem herein is modeled as a colored "optimization problem", by the definition of a " energy function " (Energy function, similar to the loss function) to formally define "problem colored artwork." When the value of the energy function reaches a minimum, to achieve optimal color results.

In order to solve this optimization problem, "energy function" need to be converted to an equivalent " Multiway Cut problem " (one kind of "Graph Cut problem" in "FIG theory"). Then provides a basis for solving the "maximum flow / minimum cost" (max-flow / min- cut) algorithm.

And "energy function" equivalent "multiway cut problem"

In this paper, the idea for solving the problem of color artwork, the main draw of the "interactive image segmentation" ( Interactive Image Segmentation , commonly known as "digging") in, Graph Cut-based segmentation method , and it was modified to suit the hand-drawn animation color. To understand the interactive image segmentation method based on Graph Cut, can be found zouxy09 of this blog .