Video compression for story-telling

These approaches are designed to let a good quality on the objects of interests while reducing it on the background. The main idea is to maintain the information necessary to take decisions. These approaches can be applied to reduce the storage of information and/or the tranmission of it. Different approaches can be done to obtain that but the main process is always the same.

The first step is to identify what is important by using peoplen car detector, saliency models... After that a reduction of the quality of the non-important area is done; The video is then encoded with a traditional encoder like H.264/AVC. The decoded video can be then post-processed to improve the semantic of the background.

Marc Decombas is working on this topic since 2010. Feel free to contact him.