公开(公告)号：US10475172B2

公开(公告)日：2019-11-12

申请号：US16017929

申请日：2018-06-25

Applicant: NETFLIX, INC.

Inventor： Anne Aaron ,  Dae Kim ,  Yu-Chieh Lin ,  David Ronca ,  Andy Schuler ,  Kuyen Tsao ,  Chi-Hao Wu

IPC: G06K9/00 ,  G06T7/00 ,  G06T9/00 ,  H04N21/466 ,  H04N19/154 ,  G06T7/20

Abstract: In one embodiment of the present invention, a quality trainer and quality calculator collaborate to establish a consistent perceptual quality metric via machine learning. In a training phase, the quality trainer leverages machine intelligence techniques to create a perceptual quality model that combines objective metrics to optimally track a subjective metric assigned during viewings of training videos. Subsequently, the quality calculator applies the perceptual quality model to values for the objective metrics for a target video, thereby generating a perceptual quality score for the target video. In this fashion, the perceptual quality model judiciously fuses the objective metrics for the target video based on the visual feedback processed during the training phase. Since the contribution of each objective metric to the perceptual quality score is determined based on empirical data, the perceptual quality score is a more accurate assessment of observed video quality than conventional objective metrics.

公开(公告)号：US10438335B2

公开(公告)日：2019-10-08

申请号：US16017929

申请日：2018-06-25

Applicant: NETFLIX, INC.

Inventor： Anne Aaron ,  Dae Kim ,  Yu-Chieh Lin ,  David Ronca ,  Andy Schuler ,  Kuyen Tsao ,  Chi-Hao Wu

IPC: G06K9/00 ,  G06T7/00 ,  G06T9/00 ,  H04N21/466 ,  H04N19/154 ,  G06T7/20

Abstract: In one embodiment of the present invention, a quality trainer and quality calculator collaborate to establish a consistent perceptual quality metric via machine learning. In a training phase, the quality trainer leverages machine intelligence techniques to create a perceptual quality model that combines objective metrics to optimally track a subjective metric assigned during viewings of training videos. Subsequently, the quality calculator applies the perceptual quality model to values for the objective metrics for a target video, thereby generating a perceptual quality score for the target video. In this fashion, the perceptual quality model judiciously fuses the objective metrics for the target video based on the visual feedback processed during the training phase. Since the contribution of each objective metric to the perceptual quality score is determined based on empirical data, the perceptual quality score is a more accurate assessment of observed video quality than conventional objective metrics.

公开(公告)号：US10404986B2

公开(公告)日：2019-09-03

申请号：US14673621

申请日：2015-03-30

Applicant: NETFLIX, Inc

Inventor： Anne Aaron ,  David Ronca ,  Ioannis Katsavounidis ,  Andy Schuler

IPC: H04N19/146 ,  H04N19/124 ,  H04N19/154 ,  H04N21/234 ,  H04N21/2343

Abstract: In one embodiment of the present invention, an encoding bitrate ladder selector tailors bitrate ladders to the complexity of source data. Upon receiving source data, a complexity analyzer configures an encoder to repeatedly encode the source data-setting a constant quantization parameter to a different value for each encode. The complexity analyzer processes the encoding results to determine an equation that relates a visual quality metric to an encoding bitrate. The bucketing unit solves this equation to estimate a bucketing bitrate at a predetermined value of the visual quality metric. Based on the bucketing bitrate, the bucketing unit assigns the source data to a complexity bucket having an associated, predetermined bitrate ladder. Advantageously, sagaciously selecting the bitrate ladder enables encoding that optimally reflects tradeoffs between quality and resources (e.g., storage and bandwidth) across a variety of source data types instead of a single, “typical” source data type.

公开(公告)号：US10007977B2

公开(公告)日：2018-06-26

申请号：US14709230

申请日：2015-05-11

Applicant: NETFLIX, INC.

Inventor： Anne Aaron ,  Dae Kim ,  Yu-Chieh Lin ,  David Ronca ,  Andy Schuler ,  Kuyen Tsao ,  Chi-Hao Wu

IPC: G06K9/00 ,  G06T7/00 ,  G06T9/00 ,  H04N21/466 ,  H04N19/154 ,  G06T7/20

CPC classification number: G06T7/0002 ,  G06T7/20 ,  G06T9/002 ,  G06T2207/10016 ,  G06T2207/20084 ,  G06T2207/30168 ,  H04N19/154 ,  H04N21/466 ,  H04N21/4666

Abstract: In one embodiment of the present invention, a quality trainer and quality calculator collaborate to establish a consistent perceptual quality metric via machine learning. In a training phase, the quality trainer leverages machine intelligence techniques to create a perceptual quality model that combines objective metrics to optimally track a subjective metric assigned during viewings of training videos. Subsequently, the quality calculator applies the perceptual quality model to values for the objective metrics for a target video, thereby generating a perceptual quality score for the target video. In this fashion, the perceptual quality model judiciously fuses the objective metrics for the target video based on the visual feedback processed during the training phase. Since the contribution of each objective metric to the perceptual quality score is determined based on empirical data, the perceptual quality score is a more accurate assessment of observed video quality than conventional objective metrics.

公开(公告)号：US20160335754A1

公开(公告)日：2016-11-17

申请号：US14709230

申请日：2015-05-11

Applicant: NETFLIX, Inc

Inventor： Anne Aaron ,  Dae Kim ,  Yu-Chieh Lin ,  David Ronca ,  Andy Schuler ,  Kuyen Tsao ,  Chi-Hao Wu

IPC: G06T7/00 ,  G06T7/20

CPC classification number: G06T7/0002 ,  G06T7/20 ,  G06T9/002 ,  G06T2207/10016 ,  G06T2207/20084 ,  G06T2207/30168 ,  H04N19/154 ,  H04N21/466 ,  H04N21/4666

Abstract: In one embodiment of the present invention, a quality trainer and quality calculator collaborate to establish a consistent perceptual quality metric via machine learning. In a training phase, the quality trainer leverages machine intelligence techniques to create a perceptual quality model that combines objective metrics to optimally track a subjective metric assigned during viewings of training videos. Subsequently, the quality calculator applies the perceptual quality model to values for the objective metrics for a target video, thereby generating a perceptual quality score for the target video. In this fashion, the perceptual quality model judiciously fuses the objective metrics for the target video based on the visual feedback processed during the training phase. Since the contribution of each objective metric to the perceptual quality score is determined based on empirical data, the perceptual quality score is a more accurate assessment of observed video quality than conventional objective metrics.

Abstract translation: 在本发明的一个实施例中，质量培训者和质量计算器通过机器学习协作建立一致的知觉质量度量。 在培训阶段，质量培训师利用机器智能技术创建感知质量模型，结合客观指标，最佳地跟踪在观看培训视频期间分配的主观度量。 随后，质量计算器将感知质量模型应用于目标视频的客观指标的值，从而生成目标视频的感知质量得分。 以这种方式，感知质量模型基于在训练阶段处理的视觉反馈，明智地融合了目标视频的客观指标。 由于每个客观度量对知觉质量得分的贡献是基于经验数据确定的，感知质量得分是比传统客观指标更准确地评估观察到的视频质量。