Google Inc.

Urvang Joshi

Debargha Mukherjee

Yue Chen

Sarah Parker

Andrew Fergus Simpson

Elliott Karpilovsky

Hui Su

Guided restoration is used to restore video data degraded from a video frame. The video frame is divided into restoration units (RUs) which each correspond to one or more blocks of the video frame. Restoration schemes are selected for each RU. The restoration schemes may indicate to use one of a plurality of neural networks trained for the guided restoration. Alternatively, the restoration schemes may indicate to use a neural network and a filter-based restoration tool. The video frame is then restored by processing each RU according to the respective selected restoration scheme. During encoding, the restored video frame is encoded to an output bitstream, and the use of the selected restoration schemes may be signaled within the output bitstream. During decoding, the restored video frame is output to an output video stream.

Guided restoration of video data using neural networks

Image coding using guided machine learning restoration may include obtaining reconstructed frame data by decoding, obtaining a restored frame by restoring the reconstructed frame, and outputting the restored frame. Obtaining the restored frame may include obtaining a reconstructed block, obtaining guide parameter values, obtaining a restored block, and including the restored block in the restored frame. Obtaining the restored block may include inputting the reconstructed block to an input layer of a trained guided convolutional neural network, wherein the neural network is constrained such that an output layer has a defined cardinality of channels, obtaining, from the output layer, neural network output channel predictions, obtaining a guided neural network prediction as a linear combination of the guide parameter values and the neural network output channel predictions, and generating the restored block using the guided neural network prediction.

Video coding with guided machine learning restoration

A super-resolution coding mode is described. An encoded image can be decoded from an encoded bitstream stored on a non-transitory computer-readable storage medium. A flag can indicate whether an image was encoded using the super-resolution mode at a first resolution. Responsive to the flag indicating that the image was encoded using the super-resolution mode, bits indicating an amount of scaling of the image are included. The image is decoded from the encoded bitstream to obtain a reconstructed image at the first resolution, and the reconstructed image is upscaled to a second resolution using the amount of scaling to obtain an upscaled reconstructed image. The second resolution is higher than the first resolution. Loop restoration parameters within the bitstream can used for look restoration filtering of the upscaled reconstructed image to obtain a loop restored image at the second resolution.

Super-resolution loop restoration

Improved transforms are used to encode and decode large video and image blocks. During encoding, a prediction residual block having a large size (e.g., larger than 32×32) is generated. The pixel values of the prediction residual block are transformed to produce transform coefficients. After determining that the transform coefficients exceed a threshold cardinality representative of a maximum transform block size (e.g., 32×32), a number of the transform coefficients are discarded such that a remaining number of transform coefficients does not exceed the threshold cardinality. A transform block is then generated using the remaining number. During decoding, after determining that the transform coefficients exceed the threshold cardinality, a number of new coefficients are added to the transform coefficients such that a total number of transform coefficients exceeds the threshold cardinality. The transform coefficients are then inverse transformed into a prediction residual block having a large size.

Transforms for large video and image blocks

A hybrid apparatus for coding a video stream includes a first encoder. The first encoder includes a neural network having at least one hidden layer, and the neural network receives source data from the video stream at a first hidden layer of the at least one hidden layer, receives side information correlated with the source data at the first hidden layer, and generates guided information using the source data and the side information. The first encoder outputs the guided information and the side information for a decoder to reconstruct the source data.

Hybrid motion-compensated neural network with side-information based video coding

Video coding may include generating, by a processor, a decoded frame by decoding a current frame from an encoded bitstream and outputting a reconstructed frame based on the decoded frame. Decoding includes identifying a current encoded block from the current frame, identifying a prediction coding model for the current block, wherein the prediction coding model is a machine learning prediction coding model from a plurality of machine learning prediction coding models, identifying reference values for decoding the current block based on the prediction coding model, obtaining prediction values based on the prediction coding model and the reference values, generating a decoded block corresponding to the current encoded block based on the prediction values, and including the decoded block in the decoded frame.

Image and video coding using machine learning prediction coding models

A super-resolution coding mode is described. Encoded image can be decoded by decoding, from an encoded bitstream, a flag indicating whether an image was encoded using the super-resolution mode. The image is encoded at a first resolution. Responsive to the flag indicating that the image was encoded using the super-resolution mode, bits indicating an amount of scaling of the image are decoded. The image is decoded from the encoded bitstream to obtain a reconstructed image at the first resolution, and the reconstructed image is upscaled to a second resolution using the amount of scaling to obtain an upscaled reconstructed image. The second resolution is higher than the first resolution. Loop restoration filtering is applied to the upscaled reconstructed image using loop restoration parameters to obtain a loop restored image at the second resolution.

Transform-level partitioning of a prediction residual block is performed to improve compression efficiency of video data. During encoding, a prediction residual block is generated responsive to prediction-level partitioning performed against a video block, a transform block partition type to use is determined based on the prediction residual block, a non-recursive transform-level partitioning is performed against the prediction residual block according to the transform block partition type, and transform blocks generated as a result of the transform-level partitioning are encoded to a bitstream. During decoding, a symbol representative of the transform block partition type used to encode transform blocks is derived from the bitstream, inverse transformed blocks are produced by inverse transforming encoded video data associated with the prediction residual block, and the prediction residual block is reproduced according to the transform block partition type and used to reconstruct the video block, which is output within an output video stream.

Extended transform partitions for video compression

Improved transforms are used to encode and decode large video and image blocks. During encoding, a prediction residual block having a large size (e.g., larger than 32&#xd7;32) is generated. The pixel values of the prediction residual block are transformed to produce transform coefficients. After determining that the transform coefficients exceed a threshold cardinality representative of a maximum transform block size (e.g., 32&#xd7;32), a number of the transform coefficients are discarded such that a remaining number of transform coefficients does not exceed the threshold cardinality. A transform block is then generated using the remaining number. During decoding, after determining that the transform coefficients exceed the threshold cardinality, a number of new coefficients are added to the transform coefficients such that a total number of transform coefficients exceeds the threshold cardinality. The transform coefficients are then inverse transformed into a prediction residual block having a large size.

An apparatus for coding a block of a frame using intra-prediction includes a memory and a processor. The processor is configured to execute instructions stored in the memory to obtain an intra-prediction mode for coding the block of the frame; select a transform type for coding a transform block of a residual block, which results from predicting the block using the intra-prediction mode; and code the transform block using the transform type. To select the transform type includes to, in a case where the intra-prediction mode is a SMOOTH_PRED, select a ADST_ADST transform type; in a case where the intra-prediction mode is a SMOOTH_H_PRED, select a DCT_ADST transform type; and in a case where the intra-prediction mode is a SMOOTH_V_PRED, select a ADST_DCT transform type.

Intra-prediction for smooth blocks in image/video

A method for encoding a block of a video stream includes generating, using pixel values of the block, block features for the block; for each candidate encoding mode of candidate encoding modes, generating, using the block features and the each candidate encoding mode as inputs to a machine-learning module, a respective encoding cost; selecting, based on the respective encoding costs, a predetermined number of the candidate encoding modes; selecting, based on the respective encoding costs of the at least some encoding modes, a best mode for encoding the block; and encoding, in a compressed bitstream, the block using the best mode.

Rate/distortion/RDcost modeling with machine learning

Coding a block of a video frame using an intra-prediction mode is disclosed. A method includes selecting first neighboring pixels, generating second pixels for use along a second edge that is opposite the first edge of the block, and generating a prediction block that includes predicted pixels. The first neighboring pixels are peripheral to the block along a first edge of the block. The second pixels are generated using third neighboring pixels that are peripheral to a third edge of the block. The third edge is different from the first edge and the second edge. Generating the prediction block includes interpolating, using a first interpolation, the first neighboring pixels and the second pixels to obtain the predicted pixels.

Systems and methods are disclosed for encoding and decoding video. For example, methods may include accessing an encoded bitstream; decoding loop restoration parameters in the encoded bitstream; after reconstruction of an image at a second resolution based on data of the encoded bitstream, upscaling the reconstructed image to obtain an upscaled reconstructed image at a first resolution, wherein the second resolution is less than the first resolution in at least one dimension; and applying loop restoration filtering to the upscaled reconstructed image using the loop restoration parameters to obtain a loop restored image at the first resolution.

Growing community of inventors

Mountain View, CA, United States of America

Urvang Joshi