Content-Aware Video Resizing

Abstract

As the technology evolves, the standard computer capabilities can now be equipped on handheld devices. With the availability of high speed computer networks, users start to get access to multimedia on various devices besides the television and personal computers. Since the screen size varies among different kinds of electronic devices, the traditional resizing methods are no longer sufficient enough to adapt the multimedia information to these devices effectively. For instance, the scaling method would distort the video contents by over adapting the original aspect ratio of the video source to another aspect ratio. Therefore, a method that can keep the aspect ratio of the video contents for any screens is currently in great demand. While there are many researches conducted recently, a general and effective solution still does not exist in the market. In this project, we propose a new method to resize the source video by taking its content importance into account based on the analysis of existing methods available and under research. The method makes use of the warping method with the assisted object tracking techniques to produce video resizing results which look like performing both local scaling on interested regions and cropping when necessary. It not only effectively preserves the aspect ratio of the interested object but keeps the overall layout of the video as well. In addition, we show its effectiveness by comparing both the quality of still image frames and the smoothness of the video motions among methods available to prove the generated results are in high quality enough to handle various situations.