Illumination Quality Assessment for Face Images: A Benchmark and a Convolutional Neural Networks Based Model

• Published in: Neural Information Processing pp. 583 - 593
• Main Authors: Zhang, Lijun, Zhang, Lin, Li, Lida
• Format: Book Chapter
• Language: English
• Published: Cham Springer International Publishing
• Series: Lecture Notes in Computer Science
• Subjects: Convolutional neural networks; Illumination quality assessment; Illumination transfer
• ISBN: 9783319700892; 3319700898
• ISSN: 0302-9743; 1611-3349
• DOI: 10.1007/978-3-319-70090-8_59

Many institutions, such as banks, usually require their customers to provide face images under proper illumination conditions. For some remote systems, a method that can automatically and objectively evaluate the illumination quality of a face image in a human-like manner is highly desired. However, few studies have been conducted in this area. To fill this research gap to some extent, we make two contributions in this paper. Firstly, in order to facilitate the study of illumination quality prediction for face images, a large-scale database, namely, Face Image Illumination Quality Database (FIIQD), is established. FIIQD contains 224,733 face images with various illumination patterns and for each image there is an associated illumination quality score. Secondly, based on deep convolutional neural networks (DCNN), a novel highly accurate model for predicting the illumination quality of face images is proposed. To make our results reproducible, the database and the source codes have been made publicly available at https://github.com/zhanglijun95/FIIQA.