resemblejs

Analyse and compare images with Javascript and HTML5. More info & Resemble.js Demo. Compatible with Node.js >8.

---

Get it

npm install resemblejs

bower install resemblejs

Example

Retrieve basic analysis on an image:

var api = resemble(fileData).onComplete(function(data) {
    console.log(data);
    /*
	{
	  red: 255,
	  green: 255,
	  blue: 255,
	  brightness: 255
	}
	*/
});

Use resemble to compare two images:

var diff = resemble(file)
    .compareTo(file2)
    .ignoreColors()
    .onComplete(function(data) {
        console.log(data);
        /*
	{
	  misMatchPercentage : 100, // %
	  isSameDimensions: true, // or false
	  dimensionDifference: { width: 0, height: -1 }, // defined if dimensions are not the same
	  getImageDataUrl: function(){}
	}
	*/
    });

Scale second image to dimensions of the first one:

//diff.useOriginalSize();
diff.scaleToSameSize();

You can also change the comparison method after the first analysis:

// diff.ignoreNothing();
// diff.ignoreColors();
// diff.ignoreAlpha();
diff.ignoreAntialiasing();

And change the output display style:

resemble.outputSettings({
    errorColor: {
        red: 255,
        green: 0,
        blue: 255
    },
    errorType: "movement",
    transparency: 0.3,
    largeImageThreshold: 1200,
    useCrossOrigin: false,
    outputDiff: true
});
// .repaint();

Note: resemble.outputSettings mutates global state, effecting all subsequent call to Resemble.

It is possible to narrow down the area of comparison, by specifying a bounding box measured in pixels from the top left:

const box = {
    left: 100,
    top: 200,
    right: 200,
    bottom: 600
};
resemble.outputSettings({ boundingBox: box });

resemble.outputSettings({ boundingBoxes: [box1, box2] });

You can also exclude part of the image from comparison, by specifying the excluded area in pixels from the top left:

const box = {
    left: 100,
    top: 200,
    right: 200,
    bottom: 600
};
resemble.outputSettings({ ignoredBox: box });

resemble.outputSettings({ ignoredBoxes: [box1, box2] });

Another way to exclude parts of the image from comparison, is using the ignoreAreasColoredWith option. Any pixels that match the specified color on a reference image will be excluded from comparison:

const color = {
    r: 255,
    g: 0,
    b: 0,
    a: 255
};
resemble.outputSettings({ ignoreAreasColoredWith: color });

By default, the comparison algorithm skips pixels when the image width or height is larger than 1200 pixels. This is there to mitigate performance issues.

You can modify this behaviour by setting the largeImageThreshold option to a different value. Set it to 0 to switch it off completely.

Resemble.js also supports Data URIs as strings:

resemble.outputSettings({ useCrossOrigin: false });
var diff = resemble("data:image/jpeg;base64,/9j/4AAQSkZJRgAB...").compareTo(
    "data:image/jpeg;base64,/9j/,/9j/4AAQSkZJRg..."
);

useCrossOrigin is true by default, you might need to set it to false if you're using Data URIs.

If you'd like resemble to return early:

resemble(img1)
    .compareTo(img2)
    .setReturnEarlyThreshold(8) // %
    .onComplete(data => {
        /* do something */
    });

Single callback api

The resemble.compare API provides a convenience function that is used as follows:

const compare = require("resemblejs").compare;

function getDiff() {
    const options = {
        // stop comparing once determined to be > 5% non-matching; this will
        // also enable compare-only mode and no output image will be rendered;
        // the combination of these results in a significant speed-up in batch processing
        returnEarlyThreshold: 5
    };

    // The parameters can be Node Buffers
    // data is the same as usual with an additional getBuffer() function
    compare(image1, image2, options, function(err, data) {
        if (err) {
            console.log("An error!");
        } else {
            console.log(data);
            /*
            {
            misMatchPercentage : 100, // %
            isSameDimensions: true, // or false
            dimensionDifference: { width: 0, height: -1 }, // defined if dimensions are not the same
            getImageDataUrl: function(){}
            }
            */
        }
    });
}

Node.js

Usage

The API under Node is the same as on the resemble.compare but promise based:

const compareImages = require("resemblejs/compareImages");
const fs = require("mz/fs");

async function getDiff() {
    const options = {
        output: {
            errorColor: {
                red: 255,
                green: 0,
                blue: 255
            },
            errorType: "movement",
            transparency: 0.3,
            largeImageThreshold: 1200,
            useCrossOrigin: false,
            outputDiff: true
        },
        scaleToSameSize: true,
        ignore: "antialiasing"
    };

    // The parameters can be Node Buffers
    // data is the same as usual with an additional getBuffer() function
    const data = await compareImages(
        await fs.readFile("./your-image-path/People.jpg"),
        await fs.readFile("./your-image-path/People2.jpg"),
        options
    );

    await fs.writeFile("./output.png", data.getBuffer());
}

getDiff();

Tests

To run the tests on Node (using Jest), type:

npm run test

There are also some in-browser tests. To run these install and run a http-server such as http-server from the root of the project. Then in the browser, navigate to localhost:8080/chai-tests/test.html, open up the developer console to see the results.

Dockerfile

For convenience I've added a simple Dockerfile to run the NodeJS tests in an Ubuntu container

docker build -t rsmbl/resemble .
docker run rsmbl/resemble

Reference to academic papers

As people have asked in the past, Resemble.js hasn't knowingly implemented any published ideas. RGBA colour comparison is simple and straightforward when working with the Canvas API. The antialiasing algorithm was developed at Huddle over several days of trial-and-error using various false-positive results from PhantomCSS tests.

---

Created by James Cryer and the Huddle development team.