Abstract: Video encoding or decoding includes performing affine motion compensation in an affine mode in which a prediction unit (“PU”) of the digital video coded in the affine mode uses inter prediction and a reference block bounding box size and determining whether the reference block bounding size exceeds a predefined threshold. In response to a determination that the reference block bounding size exceeds the predefined threshold, the affine motion compensation is performed using a first motion compensation operation. In response to a determination that the reference block bounding size does not exceed the predefined threshold, the affine motion compensation is performed using a second motion compensation operation that is different from the first motion compensation operation.

Abstract: Described herein are neural network-based systems, methods and instrumentalities associated with estimating the motion of an anatomical structure. The motion estimation may be performed using a feature pyramid and/or a motion pyramid that correspond to multiple image scales. The motion estimation may be performed using neural networks and parameters that are learned via a training process involving a student network and a teacher network pre-pretrained with abilities to apply progressive motion compensation.

Abstract: Embodiments of the present invention disclose a method of content-layer based compression of a video being broadcasted over a network. The method may include: receiving a video stream comprising a plurality of video stream frames; identifying in at least some of the plurality of video stream frames at least two content-layers of predefined content-layers to yield corresponding at least two content-layer streams, wherein each of the at least two content-layer streams is associated with one of the at least two content-layers; and compressing each of the at least two content-layer video streams according to predetermined parameters of the content-layer associated with the respective content-layer video stream and according to available resources of the network to yield corresponding at least two compressed content-layer streams.

Abstract: Different implementations are described; particularly implementations for video encoding and decoding using merge lists comprising history-based motion vector predictor candidates and averaged motion vector predictor candidates. The method comprises: determining for a current block of the picture one or more history-based motion vector predictor candidates using motion information of blocks preceding the current block; determining for the current block one or more averaged motion vector predictor candidates by averaging a pair of motion vector predictor candidates in a list of motion vector predictor candidates, wherein the motion vector predictor candidates in the list are formed using motion information of blocks spatially and/or temporally surrounding the current block; and inserting one or more of the averaged motion vector predictor candidates before one or more of the history-based motion vector predictor candidates in the list of motion vector predictor candidates.

Abstract: A prediction direction selection method includes constructing, according to texture consistency of a prediction unit relative to an upper-layer coding unit of the coding unit in which the prediction unit is located, a reference direction corresponding to the prediction unit. The method also includes obtaining, for the coding unit, preliminarily selected direction vectors that are selected from intra-frame directions according to rate-distortion costs. The method next includes determining processed candidate direction vectors by using the reference direction and the preliminarily selected direction vectors. The processed candidate direction vectors are used for selecting a direction corresponding to an optimal cost as a prediction direction for unit prediction performed for the coding unit.

Abstract: A method of irregular motion compensation includes using contours of objects in a reference image to tile the reference image into a plurality of irregular shapes, and mapping each irregular shape to a location in a target image by assigning a motion vector to each irregular shape.