Affective Real-Time Video Abstraction Using CUDA

In this paper, we present a real-time video abstraction framework that responds to the affection of the contents. For abstraction, we extract edges and quantize the color of each frame of video in real-time in order to emphasize the important part of the frame. The key of our scheme is how to process each frame in real-time. This purpose is achieved by using CUDA environment, which is the GPU programming environment supported by nVidia, Inc. Using CUDA, our scheme can process more than eight frames per second in producing affectively abstracted images.