Generating a video description file comprising both spatial and temporal information.

A method and device for generating a description file (e.g., manifest file or Media Presentation Description (MPD)) about a video sequence at a server device for a client device, to retrieve a video segment, comprises, for each video segment: determining a time interval (start frame, end frame) during which a detected object or region of interest (ROI) is spatially included in a same frame region in the video sequence; and generating a descriptor file comprising spatial information describing the frame region and temporal information describing a duration at least equal to the determined time interval (period). The position of the ROI in each frame may be preliminarily detected, using object content recognition, before video sequence encapsulation, with compression coding before or at the same time as encapsulation. An end frame may be determined based on ROI or object differences between present and previous frames. An additional time interval may be determined, in which a region of interest overlaps a frame region by a predetermined fraction amount. Scalable (base / enhancement layer) video coding (SVC), using independently decodable tiles, may be used.