Platform for video over Video placement
The BIZAR project is a Hazitek project that we carried out with Pausoka between 2017 and 2019. The main objective of the project was to develop a platform for "Video over Video placement". That is, to allow the customisation of audiovisual productions for different countries in order to facilitate the introduction of customised elements (mainly video and animations, but also images) in the original audiovisual production for broadcasting in different countries or territories.
Vicomtech carried out the development of the vision system, feature extraction of the area of interest, taking into account camera movements and occlusions. In addition, an algorithm was developed for the automatic detection of the mask that defines the image region to be replaced. This algorithm gives the possibility of not having to define the mask in each frame of the video, but only in one frame for each scene, thus reducing the manual work of video editing. Finally, a rendering tool has been developed that uses the results obtained through the vision system to render video on video and add additional effects, such as uniform lighting between the rendered mask and other parts of the image.
This model fits both broadcast content and OTT (Over The Top) operators of audiovisual content distribution on the network (such as Youtube, Netflix, or equivalent) that could customise the production for different territories or users as needed.