FUSION OF MULTISPECTRAL FACE IMAGES USING TRANSFORM BASED METHODS


Kurban T.

10TH INTERNATIONAL CONFERENCE ON IMAGE PROCESSING, WAVELET AND APPLICATIONS, Kocaeli, Türkiye, 18 - 20 Ekim 2019, ss.16-21

  • Yayın Türü: Bildiri / Tam Metin Bildiri
  • Basıldığı Şehir: Kocaeli
  • Basıldığı Ülke: Türkiye
  • Sayfa Sayıları: ss.16-21
  • Erciyes Üniversitesi Adresli: Evet

Özet

 Inspection of an object or a scene using more than one sensor and capturing images at the different wavelengths of the spectrum provide much more valuable information from the object or scene. Evaluation of the data becomes more complex while the number of spectral bands are increased, hence the idea of fusing images obtained at different wavelengths is emerged. The aim of multispectral image fusion is the combination of the information existed in different bands to enhance the complementary features. Fused image obtained by combining the images captured at two or more bands, becomes more useful for many applications such as face recognition. In this paper, face images obtained from LDHF (long distance heterogeneous face) database are fused with discrete wavelet transform, Laplacian pyramid and cross bilateral filter methods. Results are compared with edge quality (QE), spatial frequency (SF), fusion factor (FF) and variance weighted structural similarity measure (Qy) metrics. Experimental results show that, LP and DWT methods are better than CBF in terms of both objective and visual evaluations