惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

T
Tenable Blog
Last Week in AI
Last Week in AI
P
Proofpoint News Feed
Engineering at Meta
Engineering at Meta
H
Help Net Security
F
Fortinet All Blogs
MyScale Blog
MyScale Blog
宝玉的分享
宝玉的分享
让小产品的独立变现更简单 - ezindie.com
让小产品的独立变现更简单 - ezindie.com
博客园 - 司徒正美
量子位
N
Netflix TechBlog - Medium
Apple Machine Learning Research
Apple Machine Learning Research
小众软件
小众软件
Recorded Future
Recorded Future
博客园 - 三生石上(FineUI控件)
Vercel News
Vercel News
aimingoo的专栏
aimingoo的专栏
I
InfoQ
Microsoft Security Blog
Microsoft Security Blog
Scott Helme
Scott Helme
The Last Watchdog
The Last Watchdog
cs.AI updates on arXiv.org
cs.AI updates on arXiv.org
IT之家
IT之家
AI
AI
WordPress大学
WordPress大学
Security Archives - TechRepublic
Security Archives - TechRepublic
Google Online Security Blog
Google Online Security Blog
U
Unit 42
V2EX - 技术
V2EX - 技术
MongoDB | Blog
MongoDB | Blog
Schneier on Security
Schneier on Security
博客园 - Franky
H
Heimdal Security Blog
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
Jina AI
Jina AI
W
WeLiveSecurity
P
Privacy & Cybersecurity Law Blog
Cloudbric
Cloudbric
B
Blog RSS Feed
N
News | PayPal Newsroom
S
Securelist
Threat Intelligence Blog | Flashpoint
Threat Intelligence Blog | Flashpoint
I
Intezer
Hacker News - Newest:
Hacker News - Newest: "LLM"
CTFtime.org: upcoming CTF events
CTFtime.org: upcoming CTF events
博客园_首页
罗磊的独立博客
H
Hackread – Cybersecurity News, Data Breaches, AI and More
雷峰网
雷峰网

Spencer Mortensen: Articles

Email address obfuscation: What works in 2026? Approximate a circle with cubic Bézier curves The typographic scale
Draw a smooth curve through a sequence of points
Spencer Mortensen · 2025-12-04 · via Spencer Mortensen: Articles

Problem: Given a sequence of points, connect the points with a smooth path that can be drawn in an SVG file or graphics program:

It turns out there is a simple algorithm that produces a gorgeous path: the smoothest path that can be built with cubic Bézier curves.

In a moment, I’ll provide the optimal algorithm. It accepts a list of points, , as its input, and computes the control points, and , needed to draw a smooth sequence of cubic Bézier curves, , connecting all of the points in :

P 0 Q 0 R 0 S 0 = P 1 Q 1 R 1 S 1 B 0 B 1
In this diagram, the two cubic Bézier curves, and , form a smooth path connecting the three points , , . Each cubic Bézier curve, , is completely determined by its control points, , , , and —which are exactly what you need to draw the curve in an SVG or graphics program.

Now, without further ado, here is the algorithm! This is a working implementation in JavaScript:

const solver = new Solver();

const P = [

new Point(0, 0),

new Point(80, 160),

new Point(160, 80)

];

const [Q, R] = solver.getControlPoints(P);

'use strict';

function Point (x, y) {

this.x = x;

this.y = y;

}

Point.plus = function (p, q) {

return new Point(p.x + q.x, p.y + q.y);

}

Point.minus = function (p, q) {

return new Point(p.x - q.x, p.y - q.y);

}

Point.times = function (c, a) {

return new Point(c * a.x, c * a.y);

}

Point.divide = function (a, c) {

return new Point(a.x / c, a.y / c);

}

'use strict';

function Solver () {}

Solver.prototype.getControlPoints = function (P) {

const Q = this.__getQ(P);

const R = this.__getR(P, Q);

return [Q, R];

}

Solver.prototype.__getQ = function (P) {

const [a, b, c, d] = this.__getProblem(P);

return this.__getSolution(a, b, c, d);

}

Solver.prototype.__getProblem = function (P) {

const a = [];

const b = [];

const c = [];

const d = [];

const n = P.length - 1;

a[0] = 0;

b[0] = 2;

c[0] = 1;

d[0] = Point.plus(P[0], Point.times(2, P[1]));

for (let i = 1; i < n-1; ++i) {

a[i] = 1;

b[i] = 4;

c[i] = 1;

d[i] = Point.plus(Point.times(4, P[i]), Point.times(2, P[i+1]));

}

a[n-1] = 2;

b[n-1] = 7;

c[n-1] = 0;

d[n-1] = Point.plus(Point.times(8, P[n-1]), P[n]);

return [a, b, c, d];

}

Solver.prototype.__getSolution = function (a, b, c, d) {

const x = [];

const n = a.length;

for (let i = 1; i < n; ++i) {

const w = a[i] / b[i-1];

b[i] = b[i] - (w * c[i-1]);

d[i] = Point.minus(d[i], Point.times(w, d[i-1]));

}

x[n-1] = Point.divide(d[n-1], b[n-1]);

for (let i = n - 2; -1 < i; --i) {

x[i] = Point.divide(Point.minus(d[i], Point.times(c[i], x[i+1])), b[i]);

}

return x;

}

Solver.prototype.__getR = function (P, Q) {

const R = [];

const n = Q.length;

for (let i = 0; i < n - 1; ++i) {

R[i] = Point.minus(Point.times(2, P[i+1]), Q[i+1]);

}

R[n-1] = Point.divide(Point.plus(P[n], Q[n-1]), 2);

return R;

}

While the algorithm above is technically all you need, here’s an extra bit of code, so you can see the points and the smooth curve that connects them:

<!DOCTYPE html>

<html lang="en">

<head>

<meta charset="utf-8">

<title></title>

<script src="index.js" defer></script>

</head>

<body></body>

</html>

'use strict';

window.addEventListener('DOMContentLoaded', function () {

const P = [

new Point(0, 0),

new Point(80, 160),

new Point(160, 80)

];

const grapher = new Grapher();

const viewBox = grapher.getViewBox(P);

const svg = grapher.getSvg(viewBox, P);

document.body.appendChild(svg);

console.log(svg.outerHTML);

});

function Point (x, y) {

this.x = x;

this.y = y;

}

Point.plus = function (p, q) {

return new Point(p.x + q.x, p.y + q.y);

}

Point.minus = function (p, q) {

return new Point(p.x - q.x, p.y - q.y);

}

Point.times = function (c, a) {

return new Point(c * a.x, c * a.y);

}

Point.divide = function (a, c) {

return new Point(a.x / c, a.y / c);

}

function Solver () {}

Solver.prototype.getControlPoints = function (P) {

const Q = this.__getQ(P);

const R = this.__getR(P, Q);

return [Q, R];

}

Solver.prototype.__getQ = function (P) {

const [a, b, c, d] = this.__getProblem(P);

return this.__getSolution(a, b, c, d);

}

Solver.prototype.__getProblem = function (P) {

const a = [];

const b = [];

const c = [];

const d = [];

const n = P.length - 1;

a[0] = 0;

b[0] = 2;

c[0] = 1;

d[0] = Point.plus(P[0], Point.times(2, P[1]));

for (let i = 1; i < n-1; ++i) {

a[i] = 1;

b[i] = 4;

c[i] = 1;

d[i] = Point.plus(Point.times(4, P[i]), Point.times(2, P[i+1]));

}

a[n-1] = 2;

b[n-1] = 7;

c[n-1] = 0;

d[n-1] = Point.plus(Point.times(8, P[n-1]), P[n]);

return [a, b, c, d];

}

Solver.prototype.__getSolution = function (a, b, c, d) {

const x = [];

const n = a.length;

for (let i = 1; i < n; ++i) {

const w = a[i] / b[i-1];

b[i] = b[i] - (w * c[i-1]);

d[i] = Point.minus(d[i], Point.times(w, d[i-1]));

}

x[n-1] = Point.divide(d[n-1], b[n-1]);

for (let i = n - 2; -1 < i; --i) {

x[i] = Point.divide(Point.minus(d[i], Point.times(c[i], x[i+1])), b[i]);

}

return x;

}

Solver.prototype.__getR = function (P, Q) {

const R = [];

const n = Q.length;

for (let i = 0; i < n - 1; ++i) {

R[i] = Point.minus(Point.times(2, P[i+1]), Q[i+1]);

}

R[n-1] = Point.divide(Point.plus(P[n], Q[n-1]), 2);

return R;

}

function Grapher () {

this.margin = 5;

this.circleRadius = 3;

this.lineWidth = 2;

}

Grapher.prototype.getViewBox = function (points) {

const [min, max] = this.__getMinMax(points);

min.x -= this.margin;

min.y -= this.margin;

max.x += this.margin;

max.y += this.margin;

for (const p of points) {

p.x -= min.x;

p.y -= min.y;

}

return new Point(max.x - min.x, max.y - min.y);

}

Grapher.prototype.__getMinMax = function (points) {

const min = new Point(points[0].x, points[0].y);

const max = new Point(points[0].x, points[0].y);

for (const point of points) {

if (point.x < min.x) {

min.x = point.x;

} else if (max.x < point.x) {

max.x = point.x;

}

if (point.y < min.y) {

min.y = point.y;

} else if (max.y < point.y) {

max.y = point.y;

}

}

return [min, max];

}

Grapher.prototype.getSvg = function (viewBox, P) {

const parser = new DOMParser();

const svgXml = this.__getSvgXml(viewBox, P);

const fragment = parser.parseFromString(svgXml, 'image/svg+xml');

return fragment.firstChild;

}

Grapher.prototype.__getSvgXml = function (viewBox, P) {

const curveXml = this.__getCurveXml(P);

const pointsXml = this.__getPointsXml(P);

return `<svg viewBox="0 0 ${viewBox.x} ${viewBox.y}" xmlns="http://www.w3.org/2000/svg">

<style>

path {

fill: none;

stroke: #000;

stroke-width: ${this.lineWidth};

}

</style>

${curveXml}

${pointsXml}

</svg>`;

}

Grapher.prototype.__getCurveXml = function (P) {

if (2 < P.length) {

return this.__getSplineXml(P);

} else if (P.length === 2) {

return this.__getLineXml(P[0], P[1]);

} else {

return '';

}

}

Grapher.prototype.__getSplineXml = function (P) {

const solver = new Solver();

const [Q, R] = solver.getControlPoints(P);

const d = this.__getSplineD(P, Q, R);

return this.__getPathXml(d);

}

Grapher.prototype.__getSplineD = function (P, Q, R) {

const segments = [

`M ${P[0].x} ${P[0].y} c`

];

for (let i = 0; i < Q.length; ++i) {

const qx = this.__getNumberText(Q[i].x - P[i].x);

const qy = this.__getNumberText(Q[i].y - P[i].y);

const rx = this.__getNumberText(R[i].x - P[i].x);

const ry = this.__getNumberText(R[i].y - P[i].y);

const sx = this.__getNumberText(P[i+1].x - P[i].x);

const sy = this.__getNumberText(P[i+1].y - P[i].y);

segments.push(`${qx} ${qy} ${rx} ${ry} ${sx} ${sy}`);

}

return segments.join(' ');

}

Grapher.prototype.__getPathXml = function (d) {

return `<path d="${d}"/>`;

}

Grapher.prototype.__getLineXml = function (a, b) {

const d = this.__getLineD(a, b);

return this.__getPathXml(d);

}

Grapher.prototype.__getLineD = function (a, b) {

const ax = this.__getNumberText(a.x);

const ay = this.__getNumberText(a.y);

const dx = this.__getNumberText(b.x - a.x);

const dy = this.__getNumberText(b.y - a.y);

return `M ${ax} ${ay} l ${dx} ${dy}`;

}

Grapher.prototype.__getPointsXml = function (points) {

const circles = [];

for (const p of points) {

const circleXml = this.__getCircleXml(p);

circles.push(circleXml);

}

return circles.join("\n");

}

Grapher.prototype.__getCircleXml = function (center) {

const cx = this.__getNumberText(center.x);

const cy = this.__getNumberText(center.y);

const r = this.__getNumberText(this.circleRadius);

return `<circle cx="${cx}" cy="${cy}" r="${r}"/>`;

}

Grapher.prototype.__getNumberText = function (number) {

return number.toFixed(3).replace(/\.?0+$/, '');

}

Why does this work?

In this section, I’ll derive the formulas that make up the majority of the code above.

gives the coordinates of a point as it moves along the th Bézier curve from time to time , starting at and ending at .

So is the velocity of that point, and is its acceleration.

Here’s what we know so far:

Because the spline is :

Because the spline is :

Because the spline is :

This constrains every interior point along the path, but it doesn’t constraint the endpoints.

So we’ll add a constraint at the beginning:

…and at the end:

This linearizes the path near the endpoints. This is a reasonable choice if we don’t know how the curve should bend at the tips, but it’s not the only constraint we could have chosen.

Notice that we have equations, and the same number of unknowns (, , and ), so we’re ready to solve for .

The equation for the th cubic Bézier curve is:
for

We can use this equation to simplify those constraints from a moment ago:

The constraint is:

The constraint becomes (using and the constraint):

The constraint becomes (using and the , constraints):

The beginning constraint becomes (using and the constraint):

The end constraint becomes (using , , with , and with ):

Together, these equations form a system of linear equations, and it’s a type that is particularly easy to solve: The Thomas algorithm was designed to handle systems of equations like this (see the JavaScript implementation above).

After solving for , we’ll use the constraint to find . This gives us , , , and , which is everything we need to draw the smooth path connecting the points in .