How Face Recognition from Images Works (Part 2)

Spread the love

Last week we took a peek at how face recognition from images works. We noted that when you recognize a face, it is because your brain recognizes patterns such as how far apart the eyebrows are, how far the eyes are apart, how far the corners of the mouth is from the ears, etc. Your brain remembers how those parts come together to form the complete face. Upon looking at a face, your brain quickly analyzes those features and the distance between them and instantly compares that face to hundreds of others that you have stored in your memory.

These features could be from someone you are looking at in person, or from a photograph of someone. Once your brain finds a match, it gives you the name of that person, along with a wealth of information, such as “high school math teacher,” or “mother-in-law,” or “president of the United States,” or even “guy I saw on the running track last week.” All of this happens in seconds. While the process varies between different technologies, this is basically how face recognition technology works.

The steps of face recognition from images:

  1. Obtaining an Image – In the first step of facial recognition from images, an image must be obtained. This could be a picture of someone standing alone or in a crowd of people. The person may or may not be aware that they are being photographed or recorded on video. The image could be of the person looking straight into the camera, or it could be a view of the person from the side. The process works better if the person is not wearing sunglasses or a hat, but information can still be obtained from a photo of a person wearing these items.
  2. Analyzing the Data – Once an image has been obtained, the data is analyzed by the face recognition system. Just as the human brain does, the software takes note of the features of the face, such as the distance between the different parts, how wide the face is, eye color, etc. Key factors in this analyzation are factors such as how far apart the eyes are, and how far the chin is from the forehead. There are various facial landmarks that are identified by these software programs. Different systems have different landmarks. These landmarks are the key to recognizing a face.
  3. Compare the Data – The data obtained from analyzing the face is then compared with the data from hundreds, thousands, or even millions of other faces in the database for that face recognition system. Over 117 million Americans have their facial data stored in police databases. Over 412 million of them are stored by the FBI. Apps such as Facebook, Google, and even phone cameras can obtain and store face recognition data. Which database the face is compared to depends on which face recognition system is being used.
  4. Making an Identification – Once the data obtained from the image in step 1 has been compared to that of the images in a database, the face recognition software looks for a match. Once a match has been found that has the same facial landmarks as the image, the system announces that it has identified the person in the image.

All of this happens very quickly. Just as the human brain analyzes all of this data to find a match within seconds, so does the face recognition software. Often, a person’s identification has been established before the person even knows that their photo has been taken. That’s how face recognition from images works.