Abstract: Methods, devices and computer-readable media for processing a compressed video to perform an inference task are disclosed. Processing the compressed video may include selecting a subset of frame encodings of the compressed video, or zero or more modalities (RGB, motion vectors, residuals) of a frame encoding, for further processing to perform the inference task. Pre-existing motion vector and/or residual information in frame encodings of the compressed video are leveraged to adaptively and efficiently perform the inference task. In some embodiments, the inference task is an action recognition task, such as a human action recognition task.

Abstract: A minimum sampling rate and a minimum sampling curve for continuous representation of a scene are disclosed. The minimum sampling rate for light-field rendering is determined in accordance with Δ ⁢   ⁢ t max = 1 2 ⁢ K Ω v ⁢   ⁢ fh d , where K&OHgr;v accounts for a light-field signal cut-off frequency, a sampling camera resolution and an output resolution, ƒ specifies a sampling camera focal length, and hd specifies a light-field depth range.

Abstract: The invention relates to manifold mosaic hopping for image-base rendering (IBR). In one embodiment, a set of manifold mosaics is constructed. Depending on the desired application, the mosaics can be concentric mosaics, parallel mosaics, as well as other types of mosaics. Each mosaic represents a number of unique viewpoints that different from one another in only one direction. From a current camera or viewing position within the set of mosaics, a new position can be moved to, such that the movement is continuous over any one mosaic, and is discrete across the mosaics. An image is rendered by local warping, without interpolation of more than one mosaic.