公开(公告)号：US20230186435A1

公开(公告)日：2023-06-15

申请号：US17551087

申请日：2021-12-14

Applicant: NETFLIX, INC.

Inventor： Christos G. BAMPIS ,  Li-Heng CHEN ,  Aditya MAVLANKAR ,  Anush MOORTHY

IPC: G06T5/00 ,  G06T7/90 ,  G06T3/40

CPC classification number: G06T5/002 ,  G06T3/40 ,  G06T7/90 ,  G06T2207/10024 ,  G06T2207/20081

Abstract: In various embodiments, an image preprocessing application preprocesses images. To preprocess an image, the image preprocessing application executes a trained machine learning model on first data corresponding to both the image and a first set of components of a luma-chroma color space to generate first preprocessed data. The image preprocessing application executes at least a different trained machine learning model or a non-machine learning algorithm on second data corresponding to both the image and a second set of components of the luma-chroma color space to generate second preprocessed data. Subsequently, the image preprocessing application aggregates at least the first preprocessed data and the second preprocessed data to generate a preprocessed image.

公开(公告)号：US20240233076A1

公开(公告)日：2024-07-11

申请号：US18617162

申请日：2024-03-26

Applicant: NETFLIX, INC.

Inventor： Li-Heng CHEN ,  Christos G. BAMPIS ,  Zhi LI

IPC: G06T3/4046 ,  G06N3/084 ,  G06T9/00

CPC classification number: G06T3/4046 ,  G06N3/084 ,  G06T9/002

Abstract: In various embodiments, a training application trains a convolutional neural network to downsample images in a video encoding pipeline. The convolution neural network includes at least two residual blocks and is associated with a downsampling factor. The training application executes the convolutional neural network on a source image to generate a downsampled image. The training application then executes an upsampling algorithm on the downsampled image to generate a reconstructed image having the same resolution as the source image. The training application computes a reconstruction error based on the reconstructed image and the source image. The training application updates at least one parameter of the convolutional neural network based on the reconstruction error to generate a trained convolutional neural network. Advantageously, the trained convolution neural network can be implemented in a video encoding pipeline to mitigate visual quality reductions typically experienced with conventional video encoding pipelines that implement conventional downsampling techniques.

公开(公告)号：US20220198607A1

公开(公告)日：2022-06-23

申请号：US17133206

申请日：2020-12-23

Applicant: NETFLIX, INC.

Inventor： Li-Heng CHEN ,  Christos G. BAMPIS ,  Zhi LI

IPC: G06T3/40 ,  G06T9/00 ,  G06N3/08

Abstract: In various embodiments, a training application trains a convolutional neural network to downsample images in a video encoding pipeline. The convolution neural network includes at least two residual blocks and is associated with a downsampling factor. The training application executes the convolutional neural network on a source image to generate a downsampled image. The training application then executes an upsampling algorithm on the downsampled image to generate a reconstructed image having the same resolution as the source image. The training application computes a reconstruction error based on the reconstructed image and the source image. The training application updates at least one parameter of the convolutional neural network based on the reconstruction error to generate a trained convolutional neural network. Advantageously, the trained convolution neural network can be implemented in a video encoding pipeline to mitigate visual quality reductions typically experienced with conventional video encoding pipelines that implement conventional downsampling techniques.