Adaptation of Images and Videos for Different Screen Sizes

With the increasing popularity of smartphones and similar mobile devices, the demand for media to consume on the go rises. As most images and videos today are captured with HD or even higher resolutions, there is a need to adapt them in a content-aware fashion before they can be watched comfortably on screens with small sizes and varying aspect ratios. This process is called retargeting. Most distortions during this process are caused by a change of the aspect ratio. Thus, retargeting mainly focuses on adapting the aspect ratio of a video while the rest can be scaled uniformly. The main objective of this dissertation is to contribute to the modern image and video retargeting, especially regarding the potential of the seam carving operator [2]. There are still unsolved problems in this research field that should be addressed in order to improve the quality of the results or speed up the performance of the retargeting process. This dissertation presents novel algorithms that are able to retarget images, videos and stereoscopic videos while dealing with problems like the preservation of straight lines or the reduction of the required memory space and computation time. Additionally, a GPU implementation is used to achieve the retargeting of videos in real-time. Furthermore, an enhancement of face detection is presented which is able to distinguish between faces that are important for the retargeting and faces that are not. Results show that the developed techniques are suitable for the desired scenarios.