Matting is a technique used to extract a foreground object from an image, and to create a new image with a transparent background. Matting is typically achieved by using a known background or reference image, and by using statistical or mathematical models to estimate the transparency of each pixel in the foreground object. Our algorithms work using a single image and „figure out” the foreground object.

Right now there is a big demand for high quality matting algorithms, for high resolution images. We will describe further a few context where we gained expertise during the years.

Hair matting

Hair matting is a method that automatically detects the hair of an individual from a picture or video and helps the user do tedious operations from hair colorization to hair replacement or easy enhancements.

Below are some examples:

Original image
Detected hair and removed background

General matting

General matting refers to the matting technique applied on any type of image content. We provide a full resolution image to test the edge accuracy of our system.

Girl

Foreground extraction