Towards automatic placement of media objects in a personalised TV experience
Brahim Allan,Ian Kegel,Sri Harish Kalidass,Andriy Kharechko,Michael Milliken,Sally McClean,Bryan Scotney,Shuai Zhang
DOI: https://doi.org/10.1007/s00530-022-00974-y
IF: 3.9
2022-09-08
Multimedia Systems
Abstract:In this paper, we describe a Proof-of-Concept implementation of an automated system to drive one aspect of a personalised, object-based TV experience, on sports content, such as football and rugby. Our Proof-of-Concept uses a sequence of analytics processes, which can be performed offline or in real time, to allow a new media object to be automatically positioned on a multi-angle broadcast video sequence without occluding any key action, thus enabling additional graphic or video content to be used to personalise the broadcast for individual viewers. First, an object detection algorithm using a deep neural network model detects the players and ball, and its filtered output defines the region of interest within each frame of the video sequence. To avoid occlusion of key action by the media object, the remaining space within the sequence of frames is analysed to propose suitable locations. Our algorithm applies layout rules to ensure the object is placed in the best position based on broadcaster-defined templates (e.g. top-left, top-right, bottom-right and bottom-left). The output shows that our Proof-of-Concepts capable of processing video sequences with multiple camera angles and proposing the start time and duration of the media object without occluding any key action.
computer science, information systems, theory & methods