A novel correlation filter based on variational calculus

Type : Publication
Auteur(s) :  Djemel Ziou, Dayron Rizo Rodriguez, Nafaa Nacereddine, Salvatore Tabbone
Année :  2019
Domaine : Electronique
Revue : Signal Processing: Image Communication
Résumé en PDF :  (résumé en pdf)
Fulltext en PDF :  (.pdf)
Mots clés :  Correlation filter, Variational calculus, Face identification, Illumination variation, Single training image, Pose classification

Résumé : 

Correlation filters have been a popular technique for tackling image classification problems. The traditionalcriteria used to design correlation filters overlook some properties that can improve their discriminative power.Therefore, new criteria are proposed to design a novel correlation filter. Such criteria take advantage ofnegative samples, spatial information and the smoothness of the correlation output space. A closed formis derived from the criteria proposed using variational calculus. Moreover, it is shown that the resultingcorrelation filter is a bandpass filter. Experiments are conducted for face identification under illuminationvariation for a single training image per subject and head pose classification. The correlation filter proposeddelivers favorable scores when compared to other correlation filters and state-of-the-art approaches