Video Motion Analysis for Landscape Image Abstraction

Abstract

Non-photorealistic rendering technique can be powerful approach to enhance both visual coherence and immersion in immersive media. Especially, painterly rendering is one of Non- photorealistic rendering that seems similar to paintings using 2D or 3D model data and artist's painting is affected by factors such as how he observes objects. In here, we can notice from paintings that painters observe the scene in a short or long time when they draw. However, there are some limits in previous works using single image. They cannot emphasize attributes of objects such as motion. In this paper, we propose the motion analysis and object classification technique from the input movie for image abstraction.