# FFmpeg Image Diff

A [NodeJS](https://nodejs.org/) image diffing library using
[FFmpeg](https://www.ffmpeg.org/).  
Creates an image showing perceptual differences and returns
[SSIM](https://en.wikipedia.org/wiki/Structural_similarity) data.

## Contents

- [Requirements](#requirements)
- [Install](#install)
- [Usage](#usage)
- [Options](#options)
  - [ssim](#ssim)
  - [similarity](#similarity)
  - [blend](#blend)
  - [opacity](#opacity)
  - [color](#color)
- [Samples](#samples)
  - [Reference image](#reference-image)
  - [Comparison image](#comparison-image)
  - [Output image](#output-image)
- [Testing](#testing)
- [License](#license)
- [Author](#author)

## Requirements

This is a thin wrapper around [FFmpeg](https://www.ffmpeg.org/) and has no other
dependencies.

## Install

```sh
npm install ffmpeg-image-diff
```

## Usage

Displaying [SSIM](https://en.wikipedia.org/wiki/Structural_similarity) data
only:

```js
const imgDiff = require('ffmpeg-image-diff')

imgDiff(
  referenceImage, // e.g. 'test/lenna.png'
  comparisonImage // e.g. 'test/lenna-edit.png'
)
  .then(ssim => {
    // Sample output: { R: 0.990149, G: 0.991289, B: 0.990579, All: 0.990672 }
    console.log(ssim)
  })
  .catch(error => {
    console.error(error)
  })
```

Creating a differential image with default options:

```js
const imgDiff = require('ffmpeg-image-diff')

imgDiff(
  referenceImage, // e.g. 'test/lenna.png'
  comparisonImage, // e.g. 'test/lenna-edit.png'
  outputImage, // e.g. 'test/out.png', overwrites an existing file
  {
    ssim: true, // true or false
    similarity: 0.01, // 1.0 - 0.01
    blend: 1.0, // 1.0 - 0.0
    opacity: 0.1, // 1.0 - 0.0
    color: 'magenta' // magenta, yellow, cyan, red green, blue or ''
  }
)
  .then(ssim => {
    // Sample output: { R: 0.990149, G: 0.991289, B: 0.990579, All: 0.990672 }
    console.log(ssim)
  })
  .catch(error => {
    console.error(error)
  })
```

## Options

### ssim

If set to `false`, disables
[SSIM](https://en.wikipedia.org/wiki/Structural_similarity) calculation.  
_Default:_ `true`

### similarity

Defines the threshold to identify image differences.  
`0.01` matches slight differences, while `1.0` only matches stark contrast.  
_Range:_ `1.0 - 0.01`  
_Default:_ `0.01`

### blend

Blend percentage for the differential pixels.  
Higher values result in semi-transparent pixels, with a higher transparency the
more similar the pixels color is to the original color.  
_Range:_ `1.0 - 0.0`  
_Default:_ `1.0`

### opacity

Opacity of the reference image as backdrop for the differential image.  
Higher values result in a more opaque background image.  
_Range:_ `1.0 - 0.0`  
_Default:_ `0.1`

### color

The color balance of the differential pixels.  
An empty string displays the default differential pixels.  
_Set:_ `['magenta', 'yellow', 'cyan', 'red', 'green', 'blue', '']`  
_Default:_ `'magenta'`

## Samples

### Reference image

![Lenna](samples/lenna.png)

### Comparison image

![Lenna Edit](samples/lenna-edit.png)

### Output image

![Lenna Diff](samples/lenna-diff.png)

## Testing

1. Start [Docker](https://docs.docker.com/).
2. Install development dependencies:
   ```sh
   npm install
   ```
3. Run the tests:
   ```sh
   npm test
   ```

## License

Released under the [MIT license](https://opensource.org/licenses/MIT).

## Author

[Sebastian Tschan](https://blueimp.net/)