VizSynth: An engine for image video synthesis in traffic-based application

Synthesize a novel scene or object from a given image is widely recognized to be important for more and more lifestyle applications, especially in the traffic-based application, such as synthesis of shadow, snow and fog. In this paper, we address the problem of synthesize novel images. The project aims at developing algorithms and a system to 1) add virtual object to an image seamlessly and 2) change the appearance of objects according to different environment. Specifically, we aim at realizing the following functions in the system: 1) Synthesize fog in images. 2) Edit an image to add specified objects into it without notable artifacts. 3) Add water reflection effects to images. 4) Synthesize shadow in specified regions in images. 5) Synthesize a snow scene based on a given image. We implement our system in several novel and widely-used approach. The experiments and applications in real-life also show the results are satisfying.