Interdigital

Apparatuses and methods are disclosed including techniques for encoding and decoding video content. Techniques disclosed include reconstructing samples of a video frame of the video data to be encoded or decoded. The video frame is divided into full-regions and partial-regions according to a grid, where the partial-regions are located along the boundary of the video frame. Then, in-loop filtering is applied to the reconstructed samples, where for a full-region adjacent to a partial-region, the in-loop filtering comprises generating an extra processing region from reconstructed samples of the partial-region, and filtering the full-region using the extra processing region.

A method and apparatus for video encoding/decoding are provided. For a current block of an image of a video encoded using a compound intra prediction mode, a first intra predictor is obtained from among a first set of intra predictors wherein the first set of intra predictors comprising at least one intra predictor being a neural-network-based intra predictor. At least one second intra predictor is obtained from among a second set of intra predictors. A prediction is obtained for the current block by combining the first intra predictor and the at least one second intra predictor and the current block is encoded or decoded based on the prediction. In a variant, a planar mode is used as a representative of the neural-network-based intra predictor for cost calculation.

A method comprising: obtaining (30,31) a 3D Gaussian representative of a part of a 3D scene, the 3D Gaussian being based on a covariance matrix; obtaining (33) a scalar value from the 3D Gaussian; obtaining (34) a simplified representation of the 3D Gaussian wherein the covariance matrix is replaced by an element depending on the scalar value; and, signalling (35) information representative of the simplified representation of the 3D gaussian.

A decoding method is disclosed. A first intra prediction mode is decoded for a block splitted into a plurality of subpartitions. For at least one subpartition of the block, a second intra prediction mode is selected from a set of intra prediction modes surrounding the first intra prediction mode based on template analysis. The block is decoded wherein the at least one subpartition of the block is decoded using the second intra prediction mode.

A system for haptic spatialization introduces the notion of a haptic source, haptic medium and haptic listener. A haptic source is a virtual transmitter that is able to provide haptic information. A haptic medium is an object with haptic properties based on physical characteristics, to modulate the propagation of a haptic signal through the object. A haptic listener is a virtual sensor that captures haptic information. A haptic engine handles the workflow and interactions between these different elements of the system. More particularly, it determines how a haptic signal is propagated according to the spatial localization and the properties of the elements. In addition, the propagation depends on haptic layers that correspond to different contexts. Syntax elements describing the elements composing such system allow to create information representative of a scene comprising spatialized haptic effects. Such information can be provided to a rendering device in order to be played back accordingly.

In various implementations, methods and devices are disclosed that process one or more syntax elements for in-loop filtering a reconstructed image region using a neural network loop filter NNLF. The one or more syntax elements may comprise an identifier of a NNLF model type for the NNLF among a reduced list of five NNLF model types; an indication specifying a size, in log 2 minus 2, in width and height of the image region processed by the NNLF; an indication specifying a size, in log 2 minus 2, in width and height of an extension of the image region processed by the NNLF or an indication specifying a maximum number minus 1 of conditional scaling parameter candidates of the image region processed by the NNLF.

Systems, methods, and instrumentalities associated with motion compensation for video blocks. An example device may obtain a current block, wherein the current block has a height and a width. The device may determine an order for performing motion compensation based on the height and the width. The order for performing motion compensation may indicate whether to perform motion compensation for the current block in a horizontal direction prior to, or subsequent to performing motion compensation for the current block in a vertical direction. The device may obtain a plurality of reference samples by performing motion compensation for the current block in the order for performing motion compensation. The device may decode the current block based on the plurality of reference samples

A method comprising: obtaining a block of a picture; determining parameters of a compression tool, the determining comprising: obtaining (710) a first matrix and a one-dimensional array of values used in a first matrixial operation representing a linear system to solve to obtain the parameters; applying (711) a fast LDL decomposition to the first matrix, the fast LDL decomposition allowing obtaining a second matrix (L), the second matrix being lower triangular with diagonal entries equal to one, a plurality of scale and shift elements representative of a third matrix (D), the third matrix being diagonal and a fourth matrix (LT), the fourth matrix being a transpose of the second matrix; and, solving (712) the linear system by sequentially solving a second, a third and a fourth matrixial operations respectively based on the second matrix and the one-dimensional array of values, the plurality of scale and shift elements and the fourth matrix; and applying the compression tool to predict the block using the parameters.

A method of mapping an input image in an input color space into an output image in an output color space, the input color space being larger than the output color space, the method comprising:obtaining (220) an intermediate color space, the input color space being larger than the intermediate color space, the intermediate color space being larger than the output color space;obtaining (221) a dynamic color space conversion operator from an analysis of the input image;applying (222) a first mapping to the input image to obtain an intermediate image in the intermediate color space using the dynamic color space conversion operator; and,applying (223) a second mapping to the intermediate image to obtain the output image applying a static color space conversion operator.

A method is presented that comprises determining that a reference block to be padded intersects a reference picture. Responsive to the determining, gradients are analyzed inside a picture part located inside the reference picture to derive a dominant direction. A block part of the reference block located outside the reference picture is padded based on the dominant direction.