truesight

A lightweight JavaScript library visualizing color in storytelling. A set of easy to use functions is provided for generating the inverse color maps and color palettes in both images and videos.

Install

Install the truesight package from npm:

yarn add truesight

Usage

Use the desired functions and utility classes as follows:

import truesight from 'truesight';
 
truesight.popularizeImage(parameters);
 
new truesight.RGBImage(colors);

API

In this section, Flow type annotations are used to denote the types for both the input and output parameters. If you're not sure how to interpret a specific type annotation, the documentation at https://flow.org/en/docs/types/ might help you (or confuse you even more).

An error will be thrown if the input parameters object does not contain the expected properties.

Due to the asynchronous nature of image loading and video stream parsing, the API is designed to be Promise-aware. Each return value will be wrapped in a Promise object holding the final result.

Image

The image quantization API consists of following three functions:

• quantizeImage maps each color in the image to its representative color using the median cut algorithm. It returns a collection of objects containing the sourceColor: RGBColor and representativeColor: RGBColor properties. The color values in a RGBColor object are found in the channels property, or red, green and blue properties.
• reduceImage extracts the most dominant colors using the same median cut implementation as before. It returns a collection of objects containing the color: RGBColor and population: number properties (or a histogram of the most dominant colors).
• popularizeImage on the other hand, extracts the most dominant colors using the popularity algorithm. It returns the same collection type of color palette entries as the reduceImage function does.

All three functions expect a parameters object containing the imageElement: HTMLImageElement | HTMLCanvasElement | RGBImage property. Note that in Flow, the vertical bar | is used to denote the union type. The RGBImage type might come in handy, when you're working with custom image types.

You can tweak the underlying quantization algorithm by using the numberOfColors and quality properties. By doing so, you change the number of target colors to use and number of pixels to skip respectively. The default value for numberOfColors is 8, while the default value for quality is 1. Consider setting the quality property to a higher value, if image quantization is taking too long.

Example

Say you have the following HTMLImageElement somewhere in an HTML document:

<img id="2001-a-space-odyssey" src="2001-a-space-odyssey_1968.jpg" width="67.5" height="100">

You then parse the inverse color map for that image like so:

import truesight from 'truesight';
 
async function foo() {
  const inverseColorMap = await truesight.quantizeImage({
    imageElement: document.querySelector('#2001-a-space-odyssey'),
    numberOfColors: 10,
    quality: 3,
  });
 
  for (const { sourceColor, representativeColor } of inverseColorMap) {
    blur(sourceColor, representativeColor);
  }
}

Video

The video quantization functions quantizeVideo, reduceVideo and popularizeVideo are implemented by running the name-fellow image quantization functions on each frame. They return an asynchronous stream of color maps. If you're new to asynchronous iteration, the Asynchronous iteration chapter in Exploring ES2018 and ES2019 might be worth a read.

Each of the video quantization functions expects a parameters object containing the videoElement: HTMLVideoElement property. In addition to the image quantization properties mentioned before, you can alter the seconds between frames parsed by setting the secondsBetweenFrames: number property. By default, this value is 1 second.

A result object in the stream contains the index, timestamp and result properties. The index property starts at 0 and denotes the index in the stream for the parsed frame, while the timestamp property denotes the video timestamp in milliseconds. Lastly, the result property simply holds the result of image quantization.

Example

Say you have the following HTMLVideoElement showing the new Incredibles 2 trailer:

<video id="incredibles-2" src="incredibles-2_2018.mp4" width="1280" height="545"></video>

You then await the color palette for every parsed frame as follows:

import truesight from 'truesight';
 
async function bar() {
  const colorPaletteStream = await truesight.popularizeVideo({
    videoElement: document.querySelector('#incredibles-2'),
    secondsBetweenFrames: 3,
    numberOfColors: 7,
    quality: 15,
  });
 
  for await (const { index, timestamp, result } of colorPaletteStream) {
    colorize(index, timestamp, result);
  }
}