Face beautification process

In recent years, opportunities to handle high-definition images have been increasing with the spread of smartphones. In such cases, it is becoming more and more common to take pictures with smartphones, process the images, and use them. Therefore, there are some needs for a technology that is easy to use and makes people more beautiful. In this study, we will develop facial enhancement algorithms. We aim to eventually distribute them to general users.

Background of this research

This research utilizes super-resolution technology to obtain high-resolution images. When super-resolution technology is used for facial images, the sharpness of those images increase and eyes can be emphasized. However, at the same time, it can also make the wrinkles on skins be emphasized. Therefore, we have been developing wrinkle-removal processing algorithms to improve the appearance of faces.

Face beautification

The following procedure is used to obtain beautiful face images. The original face image is divided into skin and non-skin parts.

bigan-eng.png

Fig.1: Block diagram for face beautification process

bigan_result-eng.png

Fig.2: Experimental results for face beautification process

Research contents

In our laboratory, we have been studying on the following aspects of facial beautification algorithms:

  • Improving super-resolution accuracy: We aim to improve the accuracy of the super-resolution process to produce more beautiful images.
  • Improving the accuracy of skin tone detection: We have been developing algorithms those are able to perform skin tone detection automatically and more accurately.
  • Faster processing: Considering practical uses, we aim to do high-speed processing.