Community Research and Development Information Service - CORDIS

Joint EU-Chinese efforts to improve video quality

Processing and storing high-definition video requires advanced software algorithms that can handle very complicated video scenes. A new initiative has made significant headway in achieving this aim.
Joint EU-Chinese efforts to improve video quality
High-definition video is becoming much more popular, yet still poses challenges when it comes to compression and storage. While current spatiotemporal saliency algorithms help process high-definition video more efficiently, there is much room for improvement in this respect, especially when the videos are visually very complicated. The EU-funded SHIVPRO (Saliency-aware high-resolution video processing) project brought together Chinese and European academic experts in the field in order to improve saliency models.

Studying videos with complicated motion, the team investigated relevant applications and technologies such as video retargeting, salient object detection, saliency aggregation and visual scanpath prediction. This opens up several new possibilities in saliency-aware applications, including saliency manipulation in images, camera autofocus and scene understanding.

Among its key outcomes, the project team devised a new spatiotemporal saliency model based on superpixel-level trajectories that is capable of upgrading saliency detection performance in challenging videos. It also proposed a quality-guided fusion approach to integrate the pixel-level temporal saliency map with the pixel-level spatial saliency map, ultimately exceeding state-of-the-art spatiotemporal saliency models on saliency detection performance.

The team also improved retargeting performance through a more effective spatiotemporal salient object detection method that maximises saliency inside the detection window. This also involves joint cropping and scaling operations based on the detected spatiotemporal salient object regions to generate the retargeted video more efficiently. Another successful project outcome involves a new framework to predict visual scanpaths of observers when watching a visual scene in order to create better saliency models.

Overall, the results can be very useful for critical applications that require high-quality saliency maps. All these positive project outcomes have been published in relevant journals and papers as well as online. Lastly, in addition to improving the processing of high-definition video, this project has opened new collaboration initiatives in the field between China and the EU.

Related information

Keywords

High-definition video, spatiotemporal saliency, SHIVPRO, saliency-aware, video processing, scanpath
Follow us on: RSS Facebook Twitter YouTube Managed by the EU Publications Office Top