Abstract: A video delivery system generates and stores reduced bandwidth videos from source video. The system may include a track generator that executes functionality of application(s) to be used at sink devices, in which the track generator generates tracks from execution of the application(s) on source video and generates tracks having a reduced data size as compared to the source video. The track generator may execute a first instance of application functionality on the source video, which identifies region(s) of interest from the source video. The track generator further may downsample the source video according to downsampling parameters, and execute a second instance of application functionality on the downsampled video. The track generator may determine, from a comparison of outputs from the first and second instances of the application, whether the output from the second instance of application functionality is within an error tolerance of the output from the first instance of application functionality.

Abstract: Improved neural-network-based image and video coding techniques are presented, including hybrid techniques that include both tools of a host codec and neural-network-based tools. In these improved techniques, the host coding tools may include conventional video coding standards such H.266 (VVC). In an aspects, source frames may be partitioned and either host or neural-network-based tools may be selected per partition. Coding parameter decisions for a partition may be constrained based on the partitioning and coding tool selection. Rate control for host and neural network tools may be combined. Multi-stage processing of neural network output may use a checkerboard prediction pattern.

Abstract: Methods and apparatus for contextual video content adaptation are disclosed. Video content is adapted based on any number of criteria such as a target device type, viewing conditions, network conditions or various use cases, for example. A target adaptation of content may be defined for a specified video source. For example, based on receiving a request from a portable device for a live sports feed, a shortened and reduced resolution version of the live sport feed video may be defined for the portable device. The source content may be accessed and adapted (e.g., adapted temporally, spatially, etc.) and an adapted version of content generated. For example, the source content may be cropped to a particular spatial region of interest and/or reduced in length to a particular scene. The generated adaptation may be transmitted to a device in response to the request, or stored to a storage device.

Abstract: Techniques are disclosed for managing reference frames for gradual coder refresh (GDR) operation. A GDR frame may be partitioned into a plurality of units, at least one of which is coded by instantaneous decoder refresh (IDR) techniques and other(s) of which are coded by other techniques such as inter-coding. The coded GDR frame may be exchanged between an encoder and a decoder. The encoder and decoder both may decode the GDR frame. The encoder and decoder may store the IDR-coded portion of the GDR frame in a reference picture buffer in a modified frame that includes, for the other portion(s) of the GDR frame, replacement content instead of the content obtained by decoding. The modified reference frame are expected by bias prediction search operations performed on later frame toward selection of the IDR-coded content as opposed to the replacement content.

Abstract: Systems and methods are disclosed for coding pixel blocks of an input frame in which coding costs of a plurality of candidate coding modes are estimated. A coding cost of a candidate coding mode may be estimated based on noise estimate associated with the candidate coding mode. A coding mode for the input pixel block may be selected based on a comparison of the estimated coding costs of the plurality of candidate coding modes. The input pixel block may then be coded according to the selected coding mode.

Abstract: Techniques are disclosed for managing reference frames for gradual coder refresh (GDR) operation. A GDR frame may be partitioned into a plurality of units, at least one of which is coded by instantaneous decoder refresh (IDR) techniques and other(s) of which are coded by other techniques such as inter-coding. The coded GDR frame may be exchanged between an encoder and a decoder. The encoder and decoder both may decode the GDR frame. The encoder and decoder may store the IDR-coded portion of the GDR frame in a reference picture buffer in a modified frame that includes, for the other portion(s) of the GDR frame, replacement content instead of the content obtained by decoding. The modified reference frame are expected by bias prediction search operations performed on later frame toward selection of the IDR-coded content as opposed to the replacement content.

Abstract: Techniques are disclosed for coding pixel blocks of an input frame in which coding costs of a plurality of candidate coding modes are estimated, in which estimates of distortion include an estimate of noise of each respective candidate coding mode. Thus, evaluation of each coding mode may involve searching for a prediction reference for a pixel block according to the mode, and estimating coding cost of the respective candidate coding mode based at least in part on an estimate of noise associated with the respective coding mode. A coding mode for the input pixel block may be selected based on a comparison of the coding costs of the candidate coding modes, and the input pixel block may be coded according to the selected coding mode. By estimating noise associated with the candidate coding modes and, in particular, employing techniques to maintain consistent levels of noise across different frames of a coded video sequence, it is expected that perceived quality of coding will be improved.

Abstract: Methods of in-loop deblocking filter for high dynamic range (HDR) video compression are disclosed. HDR processing and standard dynamic range (SDR) processing adopt different electro-optical transfer function (EOTF) to convert digital code words to linear luminance. For compressing HDR video, EOTF is proposed to be involved in the selection of two deblocking parameters, ? and tC, which control the strength of deblocking filter. In local activity checking for filter decisions, the calculated local signal characteristics and the thresholds are adjusted according to EOTF. After deblocking filter, the clipping range is modified based on EOTF. The chroma deblocking filter is also extended to inter-inter block boundary.

Abstract: Video coding techniques are disclosed for resource-limited destination display devices. Input video data may be coded by converting a first representation of the input video to a resolution of a destination display and base layer coding the converted representation. Additionally, a region of interest may be predicted from within the input video. The predicted ROI may be converted to a resolution of the destination display, and the converted ROI may be enhancement layer coded. The base layer coded data and the enhancement layer data may be transmitted to the destination display where the coded base layer data is decoded and displayed until a zoom event occurs. When a zoom event occurs, both the coded base layer data and the coded enhancement layer data may be decoded and displayed. Thus, the switchover from a first field of view to an ROI view may be performed quickly.

Abstract: Methods of in-loop deblocking filter for high dynamic range (HDR) video compression are disclosed. HDR processing and standard dynamic range (SDR) processing adopt different electro-optical transfer function (EOTF) to convert digital code words to linear luminance. For compressing HDR video, EOTF is proposed to be involved in the selection of two deblocking parameters, ? and tC, which control the strength of deblocking filter. In local activity checking for filter decisions, the calculated local signal characteristics and the thresholds are adjusted according to EOTF. After deblocking filter, the clipping range is modified based on EOTF. The chroma deblocking filter is also extended to inter-inter block boundary.

Abstract: Video coding techniques are disclosed for resource-limited destination display devices. Input video data may be coded by converting a first representation of the input video to a resolution of a destination display and base layer coding the converted representation. Additionally, a region of interest may be predicted from within the input video. The predicted ROI may be converted to a resolution of the destination display, and the converted ROI may be enhancement layer coded. The base layer coded data and the enhancement layer data may be transmitted to the destination display where the coded base layer data is decoded and displayed until a zoom event occurs. When a zoom event occurs, both the coded base layer data and the coded enhancement layer data may be decoded and displayed. Thus, the switchover from a first field of view to an ROI view may be performed quickly.

Abstract: Methods and apparatus for contextual video content adaptation are disclosed. Video content is adapted based on any number of criteria such as a target device type, viewing conditions, network conditions or various use cases, for example. A target adaptation of content may be defined for a specified video source. For example, based on receiving a request from a portable device for a live sports feed, a shortened and reduced resolution version of the live sport feed video may be defined for the portable device. The source content may be accessed and adapted (e.g., adapted temporally, spatially, etc.) and an adapted version of content generated. For example, the source content may be cropped to a particular spatial region of interest and/or reduced in length to a particular scene. The generated adaptation may be transmitted to a device in response to the request, or stored to a storage device.

Abstract: In accordance with an embodiment, a method of watermarking encoded video frames includes electronically receiving a bitstream comprising a plurality of encoded video frames that are divided into a plurality of macro-blocks, determining macro-block dependencies based on prediction information in the received bitstream, determining a set of macro-blocks having a minimal number of macro-block dependencies, and embedding a watermark in a plurality of macro-blocks selected from the set of macro-blocks having the minimal number of macro-block dependencies.

Abstract: In accordance with an embodiment, a method of watermarking encoded video frames includes electronically receiving a bitstream comprising a plurality of encoded video frames that are divided into a plurality of macro-blocks, determining macro-block dependencies based on prediction information in the received bitstream, determining a set of macro-blocks having a minimal number of macro-block dependencies, and embedding a watermark in a plurality of macro-blocks selected from the set of macro-blocks having the minimal number of macro-block dependencies.