paper reading(1) - Combining Sketch and Tone for Pencil Drawing Production

目录

  • Combining Sketch and Tone for Pencil Drawing Production

    • paper content understanding

      • algorithm understand
      • report outlines
    • paper writing strategies note
      • Abstract
      • Introduction
      • Related Work

Combining Sketch and Tone for Pencil Drawing Production



paper content understanding

algorithm understand

variable name variable definition
\(img\) input image
\(G\) gradient map of \(img\)
\(L_{i}\) \(i_{th}\) filter of conv computation
\(S\) output line drawing part
\(J\) textual map
\(R\) output pencil drawing

\(Algorithm\quad as\quad following\)

  • line drawing with stoke

    • gradient map \(G\) of \(img\)

      • \(G = \sqrt{(\frac{\partial img}{\partial x})^{2} + (\frac{\partial img}{\partial y})^{2}}\)
    • using convolution compute to get the response map
      • \(G_{i} = L_{i} * G\)
      • \(L_{i}\) is the \(i_{th}\)convolution filter, \(*\) is convolution compute
    • save maximum of each pixel in \(G_{i}\) to \(C_{i}\)
      • \(C_{i} = if\quad (argmin_{i}G_{i} == i)\quad then\quad G_{i}\quad else\quad 0\)
    • sum up of \(C\) then get line drawing with stoke
      • \(S = \sum_{i}{L_{i} * C_{i}}\)
  • texture
    • separate \(img\) tone as 3 levels

      • high level(bright) - Laplacian distribution

        • \(p_{1}(v) = if\quad (v\le 1)\quad \frac{1}{\sigma_{b}}e^{-\frac{1 - v}{\sigma_{b}}}\quad else\quad 0\)
      • middle level - average distribution
        • \(p_{2}(v) = if\quad (u_{a}\le v \le u_{b})\quad \frac{1}{u_{a} - u_{b}}\quad else\quad 0\)
      • low level(dark) - Gaussian distribution
        • \(p_{3}(v) = \frac{1}{\sqrt{2\pi \sigma_{d}}}e^{-\frac{(v - \mu_{d})^{2}}{2\sigma_{d}^{2}}}\)
    • sum up of 3 level map above
      • \(p(v) = \frac{1}{Z}\sum_{i = 1}^{3}\omega_{i}p_{i}(v)\)
      • \(\omega_{i}\) is parameter that empower \(p_{i}(v)\)
    • \(H\) is style map of pencil drawing, fitting \(J\) using \(H\), \(H(x)^{\beta(x)} \approx J(x)\)
      • \(\beta^* = argmin_{\beta}||\beta \ln H - \ln J||_{2}^{2} + \lambda||\nabla\beta||_{2}^{2}\)
    • get pencil drawing textual map
      • \(T = H^{\beta^*}\)
  • stack 2 section above together, get output pencil drawing of \(img\)
    • \(R = S·T\)

report outlines

  • abstract
  • related work
  • our model
  • comparison
  • ending remarks

paper writing strategies note

Abstract

Para.1

  • We propose …

我们提出了…

  • … is also incorporated in

… 被添加了进来

Introduction

Para.1

  • … is one of the most fundamental … in

说明重要性

  • Pencil method general fall into two categories,

Pencil 方法大致分为两类

  • The majority of … resorts to …

… 大部分的依靠…付诸实现(诉诸于完成)

  • Consequently,

因此,

  • substantial increase

大量的增长

Para.2

  • In the literature,

在文献中,

  • ~ is effortless

~ 毫不费力, 容易做得到

  • accurately extracting and manipulating structures

精确提取和操作结构
\(e.g.\) It is because accurately extraction and manipulating structures, which is almost effortless using 3D models with unknown boundary and geometry, becomes challenging due to the existing of texture, noise, and illumination variation.

  • spurious noise-like structures

虚假的噪声结构

  • without ……, … become difficult

强调 … 的重要性

  • … risk …

… 可能会导致
\(e.g.\) Adding directional hatching patterns using closely placed parallel lines risks laying them across different surfaces and depth layers, causing unnatural drawing effects.

Para.3

  • … is paramount in … (doing)

… 在 … 中至关重要

  • on the inherent difficulty to

固有的困难

Para.4

  • differ out method from existing approaches

我们的方法与已有的方法相区别开

Related Work

原文地址:https://www.cnblogs.com/litun/p/12041475.html

时间: 2024-08-29 14:10:00

paper reading(1) - Combining Sketch and Tone for Pencil Drawing Production的相关文章

关于Cewu Lu的《Combining Sketch and Tone for Pencil Drawing Production》一文铅笔画算法的理解和笔录。

相关论文的链接:Combining Sketch and Tone for Pencil Drawing Production 第一次看<Combining Sketch and Tone for Pencil Drawing Production>一文是在两年前,随意看了一下,觉得论文里的公式比较多,以为实现有一定的难度,没有去细究,最近在作者主页上看到有 [code of direction classification] 部分代码,下载后觉得还是有自己实现的可能,下面记录下自己实现过程中

Combining Sketch and Tone for Pencil Drawing Production的优化过程

Combining Sketch and Tone for Pencil Drawing Production 是一个非常出色的自然图像转铅笔画的算法,具体的算法原理就不多说了,不了解的可以看一下这个博客,写得非常清楚了: http://blog.csdn.net/bluecol/article/details/45422763 这里主要写一下我的开发过程,和用到的优化方法. 这个算法主要有三个步骤: (1) Generate Structure Map (2) Generate Tone Ma

关于Cewu Lu等的《Combining Sketch and Tone for Pencil Drawing Production》一文铅笔画算法的理解和笔录。

相关论文的链接:Combining Sketch and Tone for Pencil Drawing Production 第一次看<Combining Sketch and Tone for Pencil Drawing Production>一文是在两年前,随意看了一下,觉得论文里的公式比较多,以为实现有一定的难度,没有去细究,最近在作者主页上看到有 [code of direction classification] 部分代码,下载后觉得还是有自己实现的可能,下面记录下自己实现过程中

基于《Combining Sketch and Tone for Pencil Drawing Production》的图像铅笔画算法的实现

一,借鉴: 本文借鉴了CSDN博主风吹夏天对此论文算法的理解:风吹夏天的图像铅笔画算法,以及香港中文大学Cewu Lu等人写的该论文的主页.原文作者和博主风吹夏天都给过代码,但是代码不全.我仔细看了原论文和该博主的文章后,基本上,大致算法思想就理通了.本文对于具体算法细节就不细述了,个人建议还是先进行原论文的研读,对该论文所进行的步骤先有个大致的了解. 二,算法思路 1,首先需要产生笔画结构 大致思路:对原图像进行梯度运算,得出大致轮廓 -----> 设计8个方向的卷积核,并依据论文中的公式对像

Paper Reading: Stereo DSO

开篇第一篇就写一个paper reading吧,用markdown+vim写东西切换中英文挺麻烦的,有些就偷懒都用英文写了. Stereo DSO: Large-Scale Direct Sparse Visual Odometry with Stereo Cameras Abstract Optimization objectives: intrinsic/extrinsic parameters of all keyframes all selected pixels' depth Inte

【Paper Reading】Bayesian Face Sketch Synthesis

Contribution: 1) Systematic interpretation to existing face sketch synthesis methods. 2) Bayesian face sketch synthesis: apply the spatial neighboring constraint to both the neighbor selection model and the wieght computation model. Problem: s代表targe

ICML 2017 Reading (1): Combining Online and Offline Knowledge in UCT

立帖要读paper,已经过了快两周了,还一篇博文都没发出来,略汗. 今天第一篇,读的是今年ICML的十年Test of Time Award得奖论文Combining Online and Offline Knowledge in UCT,来自Sylvain Gelly和David Silver,发表于十年前的ICML 2007(所以,我还是没有开始讨论今年的paper...).作者Sylvain Gelly现在Google Brain.David Silver现在在DeepMind任职,是Al

Paper Reading: Perceptual Generative Adversarial Networks for Small Object Detection

Perceptual Generative Adversarial Networks for Small Object Detection 2017-07-11  19:47:46   CVPR 2017 This paper use GAN to handle the issue of small object detection which is a very hard problem in general object detection. As shown in the followin

【Paper Reading】Object Recognition from Scale-Invariant Features

Paper: Object Recognition from Scale-Invariant Features Sorce: http://www.cs.ubc.ca/~lowe/papers/iccv99.pdf SIFT 即Scale Invariant Feature Transfrom, 尺度不变变换,由David Lowe提出.是CV最著名也最常用的特征.在图像目标识别的应用中,常常要求图像的特征有很好的roboust即不容易受到平移,旋转,尺度缩放,光照,仿射的英雄.SIFT算子具有