Learning Templates for Artistic Portrait Lighting Analysis

Learning Templates for Artistic Portrait Lighting Analysis

  Xiaowu Chen, Xin Jin1,Hongyu Wu, and Qinping Zhao

1Beijing Electronic Science and Technology Institute, Beijing 100070, China


   Lighting is a key factor in creating impressive artistic portraits. In this paper, we propose to analyse portrait lighting by learning templates of lighting styles. Inspired by the experience of artists, we first define several novel features which describe the local contrasts in various face regions. The most informative features are then selected with a stepwise feature pursuit algorithm to derive the templates of various lighting styles. After that, the matching scores which measure the similarity between a testing portrait and those templates are calculated for lighting style classification. Furthermore, we train a regression model by the subjective scores and the feature responses of a template to predict the score of a portrait lighting quality. Based on the templates, a novel Face Illumination Descriptor (FID) is defined to measure the difference between two portrait lightings. Experimental results show that the learned templates can well describe the lighting styles, while the proposed approach can assess the lighting quality of artistic portraits as human being does.


The face regions definition. (a) The artists’ common practice to analyze portrait lighting quality. The face is coarsely divided into regions [11]. (b)Inspired by the artists, we divide the face into 16 rectangular parts.

The templates of four lighting styles. The left shows part of the training portraits. The right is the learned templates of each lighting style.

Local lighting contrast type design with the help of artists. We first collect a dataset of artist lighting portraits. Then, the portraits are divided into local patches (the patches of faces are regularized to squar e). Finally, the local patches are classified by artists and the lighting contrast types are derived from the local patches.

The learned artistic lighting templates. The first row is the common template of the whole artistic lighting style and the template of Paramount style.Paramount style has no sub-classes according to key lighting directions. The second to the fourth rows show the learned templates for the other three typical artistic lighting styles. In each row, from the left to the right are the common template of this style, the samples of this style, and the two templates of the two sub-classes. All templates are shown with ranks marked by colored boundaries corresponding to the legend.

Weakly supervised clustering. All the artistic lighting styles are divided into four typical artistic lighting styles. We make weakly supervised clustering within each style. Each style of Rembrandt,Loop and Split can be automatically divided into two sub-classes: Left and Right. The paramount has no such sub-classes because the key lighting direction is always above its subject.

Definition of FID. (a) The portrait photo with facial landmarks, (b)The facial mask according to (a), (c) The 16 selected contrast types.

Paper and Slides


Xiaowu Chen, Xin Jin, Hongyu Wu, and Qinping Zhao.Learning Templates for Artistic Portrait Lighting Analysis, IEEE Transations on Image Processing-submission,  2014

[pdf] Xin Jin, Mingtian Zhao,Xiaowu Chen,Qinping Zhao, and Song-Chun Zhu.Learning Artistic Lighting Template from Portrait Photographs



  author    = {Xiaowu Chen and
               Xin Jin and
               Hongyu Wu and
               Qinping Zhao},
  title     = {Learning Templates for Artistic Portrait Lighting Analysis},
  journal   = {{IEEE} Trans. Image Processing},
  volume    = {24},
  number    = {2},
  pages     = {608--618},
  year      = {2015},
  url       = {https://doi.org/10.1109/TIP.2014.2369962},
  doi       = {10.1109/TIP.2014.2369962},
  timestamp = {Fri, 26 May 2017 22:51:41 +0200},
  biburl    = {http://dblp.uni-trier.de/rec/bib/journals/tip/ChenJWZ15},
  bibsource = {dblp computer science bibliography, http://dblp.org}

3 thoughts on “Learning Templates for Artistic Portrait Lighting Analysis

  • Pingback:My Homepage

  • 2017年12月14日 at 下午7:39

    472587 138733Excellent blog here! Also your web site loads up fast! What host are you using? Can I get your affiliate link to your host? I wish my website loaded up as quickly as yours lol 882836

  • 2017年12月30日 at 上午5:33

    I have noticed you don’t monetize your website, don’t
    waste your traffic, you can earn additional cash every month because you’ve got hi quality content.
    If you want to know how to make extra bucks, search for:
    Boorfe’s tips best adsense alternative



电子邮件地址不会被公开。 必填项已用*标注