A Comparative Study on Face Sketch Synthesis

The purpose of face sketch synthesis technique is to generate a sketch from an input face image given a set of face sketch-photo images as the training set. Since face sketch synthesis is attracting huge attentions, an experimental study to existing methods is nontrivial. This paper provides a comprehensive review and comparative study to representative face sketch synthesis methods. These methods are further sub-divided into two main categories: data-driven methods, also known as exemplar-based methods, and model-driven methods. Generally, exemplar-based face sketch synthesis consists of four parts: patch representation; neighbor selection; weight computation; and patch assembling. Model-driven methods explicitly learn the mapping from face photos to face sketches. We have drawn some promising conclusions in this paper which have not been investigated before.