1. Introduction
In many computer vision and pattern recognition applications, people often have images of the same scene but obtained from different sources, and consequently the conversion between the images of different styles are required. For example, in law enforcement we may need to compare mug-shot photos to a sketch drawn by an artist based on the verbal description of the suspect. In addition, a low resolution image/video captured by low-end devices often needs to be up-converted to a higher resolution for better visualization and interpretation. Researches on such cross-style image synthesis problems can not only benefit the practical applications (e.g., public security) but also help people understand how the human visual system perceives the distinctive information of the same scene across different sources.