nikolajbech / underwater-image-color-correction

The algorithm can color correct underwater images
87 stars 18 forks source link

underwater-image-color-correction

Installation

npm i underwater-image-color-correction

Usage

The arguments in the function is the following:

The output of the function is a color filter matrix:

[
    RedRed,     RedGreen,    RedBblue,   RedAlpha,     RedOffset,
    GreenRed,   GreenGreen,  GreenBlue,  GreenAlpha,   GOffset,
    BlueRed,    BlueGreen,   BlueBlue,   BlueAlpha,    BOffset,
    AlphaRed,   AlphaGreen,  AlphaBlue,  AlphaAlpha,   AOffset,
]

The color filter matrix can be applied by an external color filter library library or the following calculation:

for (var i = 0; i < data.length; i += 4) {
    data[i] = Math.min(255, Math.max(0, data[i] * filter[0] + data[i + 1] * filter[1] + data[i + 2] * filter[2] + filter[4] * 255)) // Red
    data[i + 1] = Math.min(255, Math.max(0, data[i + 1] * filter[6] + filter[9] * 255)) // Green
    data[i + 2] = Math.min(255, Math.max(0, data[i + 2] * filter[12] + filter[14] * 255)) // Blue
}

Test in browser: https://colorcorrection.firebaseapp.com/

Motivation

As you descent underwater some colors are absorbed more than others. The red color will disappear as the first thing which will make the image appear more blue and green.

By adjusting each color channel, the colors can be restored and make the image appear more natural with higher contrast. This algorithm aims at doing this automatically.

Examples

alt text

alt text

alt text

The algorithm

  1. Calculate the average color for image.
  2. Hueshift the colors into the red channel until a minimum red average value of 60.
  3. Create RGB histogram with new red color.
  4. Find low threshold level and high threshold level.
  5. Normalize array so threshold level is equal to 0 and threshold high is equal to 255.
  6. Create color filter matrix based on the new values.