Automatic Portrait Image Selection for Smart Phones
Erfan Entezami, Maryam Karimi · 2020
Manual selection of portrait photos that are consecutively captured is usually a tedious task. Since it is very difficult to collect private selections of good images of people, using automatic methods can be helpful in choosing photos, especially in large collections. To achieve this, we devised a method assigns one of the three labels including unacceptable condition, acceptable condition, or best condition to each portrait image. After a preprocessing and face and eye detection steps, the proposed method employs the blink detection and Iris detection units to ensure that the eyes are both open and the gaze direction is straight. According to our experimental results, this method is very acceptable both in terms of accuracy and speed, and it can be reliably embedded in mobile phones and digital cameras.