转自:https://sighack.com/post/chaikin-curves

In this post, we’ll look at what Chaikin curves are, how to implement them in Processing as well as some ways you can use them in your artwork.

If you’re looking for a copy-pastable version of this algorithm in Processing, scroll to the end of the post!

All Processing code for this article, along with images and animated GIFs, can be found on Github

Chaikin Curves

In 1974, George Chaikin presented one of the first refinement-based corner cutting algorithms to generate a curve from a given set of control points. His approach involved generating a new curve by cutting the corners off the original (based on some fixed ratio), and repeating this a bunch of times.

Here is a visual demonstration of Chaikin’s corner-cutting algorithm with three iterations and a cut ratio of 25%:

Animation of Chaikin’s Corner-Cutting Algorithm for a Polyline

Animation of Chaikin’s Corner-Cutting Algorithm for a Closed Polygon

What Can I Do With Them?

Let’s now look at some ways you can incorporate Chaikin curves into your artwork.

Here is an example of using Chaikin’s corner-cutting algorithm to round out the polygons of a Voronoi diagram. As you can see, doing this gives a nice “cobbled” effect:

And here is the same image with the original Voronoi diagram overlaid on top in red, to show you how the rounding works:

Here is another example where you can use deformed polygons (i.e. a regular polygon with some random perturbation in each vertex) as an input to the Chaikin corner-cutting algorithm to produce blob-like shapes:

Here is a similar idea to generating blobs as above, but instead overlaying them on top of one another at a low opacity to give a soft edge:

Below is a texture created from a single Chaikin curve. A random walk is used to create a sequence of line segments as we jump around the canvas, and it’s smoothed out at the end using the corner-cutting algorithm:

Here is a final example of creating a soft brush-like texture for your lines. Similar to the previous one, the image below uses a single set of points and applies a corner-cutting algorithm on it. However, I do this multiple times, with slight perturbations in the positions of the vertices each time.

As you can see, the approach is fairly versatile and there are many ways you can incorporate Chaikin curves in your artwork! Let’s now look at how to implement this simple algorithm in Processing.

Implementation in Processing

Although the code I will show here is in Processing/Java, I want you to take away the following high-level algorithmic idea so you can implement it in any language of your choice.

High-Level Algorithm

The basic procedure for generating Chaikin curves is called the corner-cutting algorithm and looks something like this:

  1. Given an initial shape (open or closed) in the form of a sequence of vertices
  2. For each interior vertex forming a corner, replace it with two new vertices representing a cut.

Let’s try to implement this algorithm to better understand how it works.

Function Skeleton

Now let’s try to implement this in Processing! We first define an empty function as shown below:

/*
* The following function takes as input a curve (in the form of
* a PShape object), a cutting ratio, the number of iterations of
* the corner-cutting algorithm to use, and a boolean specifying
* whether the PShape provided is an open shape (e.g. line) or a
* closed polygon (e.g. hexagon).
*/
PShape chaikin(PShape shape, float ratio, int iterations, boolean close) {
PShape next = createShape(); // Take the shape specified in the 'shape' variable and
// generate a Chaikin version out of it inside 'next'. return next;
}

Above, we define an empty function the takes as input a PShape object. This is a handy way to store a shape comprising multiple vertices in Processing. We can use it to specify both open shapes (e.g. lines) as well as closed ones (i.e. polygons).

Since we’re going to be cutting the edges attached to each edge based on some fixed ratio, we take a ratio parameter that has a value between zero and one. For example, a ratio of 0.25 would mean cutting each edge one quarter and three quarters in.

Next, since we can run the cutting algorithm multiple times, we accept an iterations parameter specifying this number.

Finally, since our PShape might have been open or closed, and Processing provides no easy way to determine this, we take a boolean argument that lets us specify it manually.

Based on this, we can define two convenience functions to deal with open and closed shapes. The only difference in the functions below is the value of the last parameter passed to the chaikin() function call.

PShape chaikin_closed(PShape shape, float ratio, int iterations) {
return chaikin(shape, ratio, iterations, true);
} PShape chaikin_open(PShape shape, float ratio, int iterations) {
return chaikin(shape, ratio, iterations, false);
}

A Recursive Algorithm

Now let’s get into the meat of it: how do we actually implement the corner-cutting algorithm?

Remember that Chaikin’s algorithm involves repeating the procedure a bunch of times (as specified by our iterations parameter). We can use recursion to achieve this; that is, we will call the chaikin() function from inside the chaikin() function itself:

PShape chaikin(PShape shape, float ratio, int iterations, boolean close) {
// If the number of iterations is zero, return shape as is
if (iterations == )
return shape; PShape next = createShape();
...
// Perform one iteration of the corner-cutting algorithm here.
... /*
* Call the chaikin() function with one less iteration and our
* newly-created shape so it's used as the new base shape. This
* returns a PShape which we give back to the user.
*
*/
return chaikin(next, ratio, iterations - , close);
}

The above code is an example of a recursive call. Let’s ignore the first if statement for now and look below that. When we first invoke the chaikin() function in our code, we first create a new shape and apply one iteration of the corner-cutting algorithm.

Now we use this new shape as the basis for the next iteration! We return the result of a chaikin() function call, but pass it the new shape as a base and give tell it to perform one less iteration (since we already performed one).

This way, the shape gets refined over and over! However, we need to stop at some point and return the shape as it currently stands. This is done when we see that the chaikin() function was invoked with an iterations value of zero. This brings us to the first if statement: we check if the number of iterations hit zero and return the shape immediately (without performing another iteration).

Cutting a Corner

Let’s now leave our skeleton function as is for now and write a function to cut a single corner of our shape:

ArrayList<PVector> chaikin_cut(PVector a, PVector b, float ratio) {
float x, y;
ArrayList<PVector> n = new ArrayList<PVector>(); /*
* If ratio is greater than 0.5 flip it so we avoid cutting across
* the midpoint of the line.
*/
if (ratio > 0.5) ratio = - ratio; /* Find point at a given ratio going from A to B */
x = lerp(a.x, b.x, ratio);
y = lerp(a.y, b.y, ratio);
n.add(new PVector(x, y)); /* Find point at a given ratio going from B to A */
x = lerp(b.x, a.x, ratio);
y = lerp(b.y, a.y, ratio);
n.add(new PVector(x, y)); return n;
}

Above, the function takes two vertices (as PVector’s) that represent a single edge of our shape, and a cut ratio between zero and one as seen before. From this, we derive two new points: one at the specified ratio when going from point A to point B, and one when going in the reverse direction. For example, cutting an edge with a ratio of 0.25 would give us two points: the first at 25% from point A and the other at 75% from point A (or 25% from point B).

You’ll notice an if condition that checks the value of ratio and inverts it if the value is greater than 0.5. This is required to avoid cutting across the midpoint of the edge, which would otherwise lead to an incorrect ordering of vertices in the final shape. This little if statement fixes that before we do all our calculations.

To calculate this, we use Processing’s lerp() function to interpolate the X and Y coordinates as shown above.

Finally, we add the two new points into an ArrayList and return it.

Drawing Into a PShape Object

Let’s now get back to the function we left off earlier and finish it up. Let’s first look at how we can draw inside a PShape in Processing:

PShape chaikin(PShape shape, float ratio, int iterations, boolean close) {
...
PShape next = createShape();
next.beginShape(); // Draw the next iteration of the shape using corner cutting if (close)
next.endShape(CLOSE);
else
next.endShape(); return chaikin(next, ratio, iterations - , close);
}

In Processing, in order to draw inside a PShape object, we must first call its beginShape() function. Once all drawing operations have completed, we must call endShape(). For closed polygons, we must additionally pass the CLOSE parameter to endShape().

Above, we beginShape() after calling createShape(), and depending on whether we specified this to be an open or closed shape, call the appropriate endShape() version (i.e. with or without the CLOSE parameter).

Cutting Corners

Now let’s look at the main logic for creating the next iteration of a given shape using the corner-cutting algorithm:

PShape chaikin(PShape shape, float ratio, int iterations, boolean close) {
...
next.beginShape(); /*
* Step 1: Figure out how many corners the shape has
* depending on whether it's open or closed.
*/
int num_corners = shape.getVertexCount();
if (!close)
num_corners = shape.getVertexCount() - ; /*
* Step 2: Since we don't have access to edges directly
* with a PShape object, do a pairwise iteration
* over vertices instead. Same thing.
*/
for (int i = ; i < num_corners; i++) { // Get the i'th and (i+1)'th vertex to work on that edge.
PVector a = shape.getVertex(i);
PVector b = shape.getVertex((i + ) % shape.getVertexCount()); // Step 3: Break it using our chaikin_cut() function
ArrayList<PVector> n = chaikin_cut(a, b, ratio); /*
* Now we have to deal with one corner case. In the case
* of open shapes, the first and last endpoints shouldn't
* be moved. However, in the case of closed shapes, we
* cut all edges on both ends.
*/
if (!close && i == ) {
// For the first point of open shapes, ignore vertex A
next.vertex(a.x, a.y);
next.vertex(n.get().x, n.get().y);
} else if (!close && i == num_corners - ) {
// For the last point of open shapes, ignore vertex B
next.vertex(n.get().x, n.get().y);
next.vertex(b.x, b.y);
} else {
// For all other cases (i.e. interior edges of open
// shapes or edges of closed shapes), add both vertices
// returned by our chaikin_cut() method
next.vertex(n.get().x, n.get().y);
next.vertex(n.get().x, n.get().y);
}
} if (close)
next.endShape(CLOSE);
else
next.endShape(); return chaikin(next, ratio, iterations - , close);
}

The logic above can be broken into three steps. We first figure out how many corners the shape has depending on whether it’s open or closed. In closed shapes (e.g., rectangles or triangles) each vertex represents a corner, making the number of vertices is the same as the number of corners in it. In the case of open shapes, the number of edges is one less that the number of vertices (since the last vertex is not connected back to the first one).

The second step is to iterate over each edge. Unfortunately, Processing doesn’t provide any function to do this directly, so we iterate over vertices in a pairwise fashion instead. Above, we get the i^th and (i+1)^th vertex in each loop iteration. We also make sure, that for closed shapes, when we reach the end, we wrap around to pick the first vertex (using the modulo operator).

Finally, for each edge, we break it using out chaikin_cut() function. This returns two vertices by cutting off both ends of our edge. At this point we deal with one last corner case: we make sure that for open shapes, we don’t change the first and last vertices and keep the originals.

And that’s it!

Here’s our final implementation in a nice copy-pasteable snippet:

ArrayList<PVector> chaikin_cut(PVector a, PVector b, float ratio) {
float x, y;
ArrayList<PVector> n = new ArrayList<PVector>(); /*
* If ratio is greater than 0.5 flip it so we avoid cutting across
* the midpoint of the line.
*/
if (ratio > 0.5) ratio = - ratio; /* Find point at a given ratio going from A to B */
x = lerp(a.x, b.x, ratio);
y = lerp(a.y, b.y, ratio);
n.add(new PVector(x, y)); /* Find point at a given ratio going from B to A */
x = lerp(b.x, a.x, ratio);
y = lerp(b.y, a.y, ratio);
n.add(new PVector(x, y)); return n;
} PShape chaikin(PShape shape, float ratio, int iterations, boolean close) {
// If the number of iterations is zero, return shape as is
if (iterations == )
return shape; PShape next = createShape();
next.beginShape(); /*
* Step 1: Figure out how many corners the shape has
* depending on whether it's open or closed.
*/
int num_corners = shape.getVertexCount();
if (!close)
num_corners = shape.getVertexCount() - ; /*
* Step 2: Since we don't have access to edges directly
* with a PShape object, do a pairwise iteration
* over vertices instead. Same thing.
*/
for (int i = ; i < num_corners; i++) { // Get the i'th and (i+1)'th vertex to work on that edge.
PVector a = shape.getVertex(i);
PVector b = shape.getVertex((i + ) % shape.getVertexCount()); // Step 3: Break it using our chaikin_break() function
ArrayList<PVector> n = chaikin_cut(a, b, ratio); /*
* Now we have to deal with one corner case. In the case
* of open shapes, the first and last endpoints shouldn't
* be moved. However, in the case of closed shapes, we
* cut all edges on both ends.
*/
if (!close && i == ) {
// For the first point of open shapes, ignore vertex A
next.vertex(a.x, a.y);
next.vertex(n.get().x, n.get().y);
} else if (!close && i == num_corners - ) {
// For the last point of open shapes, ignore vertex B
next.vertex(n.get().x, n.get().y);
next.vertex(b.x, b.y);
} else {
// For all other cases (i.e. interior edges of open
// shapes or edges of closed shapes), add both vertices
// returned by our chaikin_break() method
next.vertex(n.get().x, n.get().y);
next.vertex(n.get().x, n.get().y);
}
} if (close)
next.endShape(CLOSE);
else
next.endShape(); return chaikin(next, ratio, iterations - , close);
} PShape chaikin_close(PShape original, float ratio, int iterations) {
return chaikin(original, ratio, iterations, true);
} PShape chaikin_open(PShape original, float ratio, int iterations) {
return chaikin(original, ratio, iterations, false);
}

Chaikin Curves in Processing的更多相关文章

  1. Video processing systems and methods

    BACKGROUND The present invention relates to video processing systems. Advances in imaging technology ...

  2. Image Processing and Analysis_15_Image Registration:Image matching as a diffusion process: An analogy with Maxwell's demons——1998

    此主要讨论图像处理与分析.虽然计算机视觉部分的有些内容比如特 征提取等也可以归结到图像分析中来,但鉴于它们与计算机视觉的紧密联系,以 及它们的出处,没有把它们纳入到图像处理与分析中来.同样,这里面也有 ...

  3. Image Processing and Analysis_8_Edge Detection:Edge and line oriented contour detection State of the art ——2011

    此主要讨论图像处理与分析.虽然计算机视觉部分的有些内容比如特 征提取等也可以归结到图像分析中来,但鉴于它们与计算机视觉的紧密联系,以 及它们的出处,没有把它们纳入到图像处理与分析中来.同样,这里面也有 ...

  4. Image Processing and Analysis_8_Edge Detection:Learning to Detect Natural Image Boundaries Using Local Brightness, Color, and Texture Cues ——2004

    此主要讨论图像处理与分析.虽然计算机视觉部分的有些内容比如特 征提取等也可以归结到图像分析中来,但鉴于它们与计算机视觉的紧密联系,以 及它们的出处,没有把它们纳入到图像处理与分析中来.同样,这里面也有 ...

  5. Image Processing and Analysis_8_Edge Detection:Statistical edge detection_ learning and evaluating edge cues——2003

    此主要讨论图像处理与分析.虽然计算机视觉部分的有些内容比如特 征提取等也可以归结到图像分析中来,但鉴于它们与计算机视觉的紧密联系,以 及它们的出处,没有把它们纳入到图像处理与分析中来.同样,这里面也有 ...

  6. Image Processing and Analysis_8_Edge Detection: Optimal edge detection in two-dimensional images ——1996

    此主要讨论图像处理与分析.虽然计算机视觉部分的有些内容比如特 征提取等也可以归结到图像分析中来,但鉴于它们与计算机视觉的紧密联系,以 及它们的出处,没有把它们纳入到图像处理与分析中来.同样,这里面也有 ...

  7. Image Processing and Computer Vision_Review:Local Invariant Feature Detectors: A Survey——2007.11

    翻译 局部不变特征探测器:一项调查 摘要 -在本次调查中,我们概述了不变兴趣点探测器,它们如何随着时间的推移而发展,它们如何工作,以及它们各自的优点和缺点.我们首先定义理想局部特征检测器的属性.接下来 ...

  8. OLTP(on-line transaction processing)与OLAP(On-Line Analytical Processing)

    OLTP与OLAP的介绍 数据处理大致可以分成两大类:联机事务处理OLTP(on-line transaction processing).联机分析处理OLAP(On-Line Analytical ...

  9. ECC-Elliptic Curves Cryptography,椭圆曲线密码编码学

    ECC ECC-Elliptic Curves Cryptography,椭圆曲线密码编码学,是目前已知的公钥体制中,对每比特所提供加密强度最高的一种体制.在软件注册保护方面起到很大的作用,一般的序列 ...

随机推荐

  1. go语言 函数return值的几种情况

    分三种情况 (以下 “指定返回值”这句话, 仅指return后面直接跟着的返回值) 退出执行,不指定返回值 (1) 函数没有返回值 package main import ( "fmt&qu ...

  2. Python 检查代码占用内存 工具和模块

    只介绍简单的使用, 更多使用方法请查看官方文档 tracemalloc 官方文档 tracemalloc文档地址 使用 import tracemalloc tracemalloc.start() # ...

  3. Unicode-objects must be encoded before hashing 错误解决办法

    提交注册用户数据后出来这个,错误原因是update()必须指定要加密的字符串的字符编码 #encryptions1 = sha1()s1.update(upwd.encode("utf8&q ...

  4. 怎么把ubuntu升级到最新版本

    首先是Ctrl+Alt+T 打开终端,然后在终端中输入指令(更新资源) sudo apt-get update 接着是对软件进行升级.(这是一个漫长的过程,需要下载资源) sudo apt-get u ...

  5. A1139 | 玩成模拟题的DFS

    考试的时候有思路了,但是没写完.这题起码要40min写,思路太诡异了. 刚刚写了一段,只过了一个case,得了18分,还行.明日再战. #include <stdio.h> #includ ...

  6. 59: loj #10215

    $des$ https://loj.ac/problem/10215 $sol$ exgcd检查 $code$ #include <iostream> #include <cstdl ...

  7. uniapp 组件传参

    父组件 <v-sub @returnDate=returnDate :backGround=backGround></v-sub> import vSub from " ...

  8. K8s Service原理介绍

    Service的工作方式有三种: 第一种: 是Userspace方式 如下图描述, Client Pod要访问Server Pod时,它先将请求发给本机内核空间中的service规则,由它再将请求, ...

  9. 分析WordPress数据表之分类标签表(功能篇)

    数据表分析 wp_terms(分类标签表) 表字段如下:term_id(分类或标签ID)name(分类或标签名)slug(别名)term_group(分类标签组,我想应该是用于排序分组) wp_ter ...

  10. 暑假gosh计划

    [要参与的事项]: 1.大创 2.CTF 3.ACM 4.自己的巴拉巴拉巴 [基本目标]: 1.大创 学完一本Java入门教材 学习Material Design,了解典型交互,进行ui初步设计 2. ...