Abstract: Techniques related to video coding include content adaptive quantization that provides a selection between objective quality and subjective quality delta QP offsets. An adaptive method generates an objective quality delta QP offset that achieves a best peak signal-to-noise ratio (PSNR) and/or structural similarity (SSIM) score, which refers to a similarity between images. Also, the adaptive method generates a subjective quality delta QP offset that achieves the best video multi-method assessment fusion (VMAF) score and/or multi-scale structural similarity (MSSIM) score.

Abstract: Techniques related to adaptive quantization matrix selection using machine learning for video coding are discussed. Such techniques include applying a machine learning model to generate an estimated quantization parameter for a frame and selecting a set of quantization matrices for encode of the frame from a number of sets of quantization matrices based on the estimated quantization parameter.

Abstract: Techniques related to quantization parameter estimation for coding intra and scene change frames are discussed. Such techniques include generating features based on an intra or scene change frame including a proportion of smooth blocks and one or both of a measure of block variance and a prediction distortion, and applying a machine learning model to generate an estimated quantization parameter for encoding the intra or scene change frame.

Abstract: Techniques related to adaptive quality boosting for low latency video coding are discussed. Such techniques include segmenting each of a number of temporally adjacent video frames into unique high encode quality regions and encoding each of the video frames by applying a coding quality boost to the high encode quality regions relative to other regions of the video frames.

Abstract: Disclosed in the embodiments of the present disclosure are a video generation method, an apparatus, an electronic device and a storage medium. The method comprises: during an image editing process of subjecting an initial image to a series of image editing operations to obtain a target image, acquiring initial key frames, the initial key frames being intermediate images between the initial image and the target image during the image editing process, the different initial key frames being intermediate images generated when different image editing operations of the series of image editing operations are triggered during the image editing process, and the intermediate images showing editing effects corresponding to the image editing operations; and generating a target video according to the initial key frames so as to record the image editing process of the target image. The present disclosure can be used for recording an image editing process.

Abstract: An example apparatus for adaptively encoding video frames includes a network analyzer to predict an instant bitrate based on channel throughput feedback received from a network. The apparatus also includes a content analyzer to generate ladder info based on a received frame. The apparatus further includes an adaptive decision executer to determine a frame rate, a video resolution, and a target frame size based on the predicted instant bitrate and the ladder outputs. The apparatus further includes an encoder to encode the frame based on the frame rate, the video resolution, and the target frame size.

Abstract: Techniques related to video coding include content adaptive quantization that provides a selection between objective quality and subjective quality delta QP offsets.

Abstract: Techniques related to adaptive quality boosting for low latency video coding are discussed. Such techniques include segmenting each of a number of temporally adjacent video frames into unique high encode quality regions and encoding each of the video frames by applying a coding quality boost to the high encode quality regions relative to other regions of the video frames.

Abstract: An example apparatus for encoding video frames includes a mask selector to select a subset of visual masks according to an actual target compression ratio and GOP configuration and a complexity estimator to estimate a picture level spatial/temporal complexity for a current frame. The example apparatus further includes a GOP adaptive visual mask selector to specify a visual mask from the subset of the visual masks corresponding to the estimated spatial and temporal complexity value a good enough picture QP deriver to derive a good enough picture QP value using the visual mask. The example apparatus also includes an adjustor to adjust the good enough picture QP value based on block level human visual system sensitivity and statistics of already encoded frames to obtain a final human visual system QP map.

Abstract: Using a fixed group of pictures (GOP) size in video encoding significantly hinders compression efficiency due to its inability to adapt to the dynamic nature of video content. While encoding leverages spatio-temporal redundancy within a GOP for compression, a predetermined size fails to capture the varying complexity of scenes. This leads to wasted bits in low-motion segments and insufficient reference frame variation for high-motion areas, resulting in visual artifacts and reduced compression efficiency. To address this limitation, a GOP size recommendation engine involving machine learning models can determine frame-level GOP size recommendations based on pre-encoder frame statistics. The frame-level GOP size recommendations are used to adapt the GOP size for encoding video frames.

Abstract: Techniques related to quantization parameter estimation for coding intra and scene change frames are discussed. Such techniques include generating features based on an intra or scene change frame including a proportion of smooth blocks and one or both of a measure of block variance and a prediction distortion, and applying a machine learning model to generate an estimated quantization parameter for encoding the intra or scene change frame.

Abstract: An embodiment may include a display processor, memory to store a 2D frame corresponding to a projection from a 360 video, and a quality selector to select a quality factor for a block of the 2D frame based on quality information from neighboring blocks of the 2D frame, including blocks which are neighboring only in the 360 video space. The system may also include a range adjuster to adjust a search range for the 2D frame based on a search area of the 2D frame, a viewport manager to determine if a request for a viewport of the 2D frame extends beyond a first edge of the 2D frame and to fill the requested viewport with wrap-around image information, and/or a motion estimator to estimate motion information based on both color information and depth information. Other embodiments are disclosed and claimed.

Abstract: In some compression techniques, in-loop filters can be included to effectively remove coding artifacts and improve the objective quality measurement at the same time. To avoid increasing complexity in the encoder, a single pass encoding solution can be implemented to efficiently and effectively make reasonably optimal in-loop filtering decisions. The solution can improve the in-loop filtering bit usage (e.g., reduce bitrate) and reduce the complexity in the encoder at the same time.

Abstract: A lightweight but effective adaptive coding tool selection system with content classification can be implemented to reduce complexity and maintain quality in a video encoder. Content classification may classify a current frame between at least three classifications: screen content, weak screen content, and natural content. Content classification may make use of two statistics, e.g., color number and variance, of blocks that are 8×8 pixels or larger in size. The statistics may be used to calculate three frame-level statistics, e.g., proportion/percentage of blocks with few colors, proportion/percentage of blocks with zero variance, and proportion/percentage of blocks with big/large variance. The frame-level statistics are used to classify the current frame. Based on the classification, coding tool control flags or control signals may be generated accordingly to configure the encoder to, e.g., turn on or off certain coding tools, and/or use certain parameter values for the coding tools.

Abstract: Different approaches for reducing complexity and computations in inter-prediction encoding are described. The approaches may involve one or more of quantization parameter and motion information being used to make a precision decision at a picture level that can improve compression efficiency. The approaches may involve finding prediction costs for the motion vector difference candidates and then performing rate-distortion optimization using the selected motion vector difference candidate having the lowest prediction cost. Prediction costs may be determined using sums of absolute transformed differences, which can be calculated efficiently in hardware. The rate-distortion cost of using merge mode with motion vector difference with the selected motion vector difference candidate may be compared against one or more other rate-distortion costs. In addition, in some scenarios, the approaches may involve finding prediction costs for a subset of the motion vector difference candidates to further reduce computations.

Abstract: Techniques related to accelerated video enhancement using deep learning selectively applied based on video codec information are discussed. Such techniques include applying a deep learning video enhancement network selectively to decoded non-skip blocks that are in low quantization parameter frames, bypassing the deep learning network for decoded skip blocks in low quantization parameter frames, and applying non-deep learning video enhancement to high quantization parameter frames.

Abstract: An example apparatus for encoding video frames includes a mask selector to select a subset of visual masks according to an actual target compression ratio and GOP configuration and a complexity estimator to estimate a picture level spatial/temporal complexity for a current frame. The example apparatus further includes a GOP adaptive visual mask selector to specify a visual mask from the subset of the visual masks corresponding to the estimated spatial and temporal complexity value a good enough picture QP deriver to derive a good enough picture QP value using the visual mask. The example apparatus also includes an adjustor to adjust the good enough picture QP value based on block level human visual system sensitivity and statistics of already encoded frames to obtain a final human visual system QP map.

Abstract: The disclosure relates to tuning configuration parameters for graphics pipeline for better user experience. A device for graphics processing, comprising: hardware engines; a graphics pipeline at least partly implemented by the hardware engines; and a tuner, coupled to the hardware engines and the graphics pipeline, the tuner to: collect statuses of the device during runtime for a previous frame; determine configuration parameters based on the collected statuses, the configuration parameters associated with three-dimensional 3D rendering, pre-processing and video encoding of the graphics pipeline; and tune the graphics pipeline with the determined configuration parameters for processing a next frame.

Abstract: Methods, articles, and systems of image processing comprise obtaining image data of frames of a video sequence. The method also includes determining multiple reference frames of a current frame in the video sequence. The multiple reference frames each have at least one motion compensated (MC) block of image data. Also, the method then includes generating a weight that factors noise, distortion variance, and dispersion distribution between the same MC block position and the current block. Thereafter, the method includes generating denoised filtered image data comprising applying one of the weights to the image data of the motion compensated (MC) block.

Abstract: Techniques related to accelerated video enhancement using deep learning selectively applied based on video codec information are discussed. Such techniques include applying a deep learning video enhancement network selectively to decoded non-skip blocks that are in low quantization parameter frames, bypassing the deep learning network for decoded skip blocks in low quantization parameter frames, and applying non-deep learning video enhancement to high quantization parameter frames.