Effect of Photometric Preprocessing on Thermal - Visible Face Recognition Performance


Torunoglu O., Saraydemir S., EDİZKAN R.

24th Signal Processing and Communication Application Conference (SIU), Zonguldak, Türkiye, 16 - 19 Mayıs 2016, ss.393-396 identifier identifier

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Cilt numarası:
  • Doi Numarası: 10.1109/siu.2016.7495760
  • Basıldığı Şehir: Zonguldak
  • Basıldığı Ülke: Türkiye
  • Sayfa Sayıları: ss.393-396

Özet

In the absence of lighting, due to the difference in the modality, it is very difficult to match the visible face database with the face images obtained in the thermal wavelength. Applying photometric preprocessing to reduce the modality difference, increases the face recognition performance from thermal to visible. In this study, different photometric preprocessing methods that can be used for face recognition problem between the images in the visible and thermal images were analyzed. Photometric methods have been tested in two different databases and Rank-k successes of the methods were compared. When Weberface normalization is used instead of DOG photometric preprocessing widely used in the literature, Rank-1 success has increased by 2-3%.