Imax Corporation

Ssimwave Inc.

Jiheng Wang

Kai Zeng

Zhou Wang

Hojatollah Yeganeh

Ahmed Badr

Abdul Rehman

Peter Olijnyk

Ian Wormsbecker

Kaiwen Ye

Christopher Kulbakas

Christopher Vytautas Olekas

Vladimir Sireaev

A method or system is disclosed to assess transmission errors in a visual media input. Domain knowledge is obtained from the visual media input by content analysis, codec analysis, distortion analysis, and human visual system modeling. The visual media input is divided into partitions, which are passed into deep neural networks (DNNs). The DNN outputs of all partitions are combined with the guidance of domain knowledge to produce an assessment of the transmission error. In one or more illustrative examples, transmission error assessment at a plurality of monitoring points in a visual media communication system is collected and assessed, followed by quality control processes and statistical performance assessment on the stability of the visual communication system.

Automatic visual media transmission error assessment

No-reference (NR) quality assessment (VQA) of a test visual media input encoding media content is provided. The test video visual media input is decomposed into multiple-channel representations. Domain knowledge is obtained by performing content analysis, distortion analysis, human visual system (HVS) modeling, and/or viewing device analysis. The multiple-channel representations are passed into deep neural networks (DNNs) producing DNN outputs. The DNN outputs are combined using domain knowledge to produce an overall quality score of the test visual media input.

No-reference visual media assessment combining deep neural networks and models of human visual system and video content/distortion analysis

Correlated quality-of-experience (QoE) and latency measures are generated at a plurality of monitoring points along a multimedia delivery chain including multiple video operations. At each of the plurality of monitoring points, an absolute QoE measure defined on a human perceptual quality scale for media content is computed, and one or more of content extraction or feature extraction on the media content are performed. To a common middleware from each of the plurality of monitoring points, the respective QoE measure and results of the one or more of content extraction or feature extraction are transmitted. Computing and updating an absolute QoE measure for each of the plurality of monitoring points is performed. Computing and updating latencies between multiple monitoring points at the middleware using the results from each of the plurality of monitoring points is also performed.

Unified end-to-end quality and latency measurement, optimization and management in multimedia communications

Detection of scaling of compressed videos or images is provided. A frequency domain transformation is applied along both horizontal and vertical directions of input video or images to generate frequency domain data. Statistics in the frequency domain data are computed for each of the horizontal and vertical directions to extract features. The features are modeled to scores along each of the horizontal and vertical directions. An original resolution of the input video or images in the horizontal and vertical directions is identified according to the scores.

Scaling factor detection for compressed images and videos

Frames of a video stream collected at first and second points along a video delivery chain are buffered into first and second buffers, respectively, the second point being downstream the video delivery chain from the first point. A control group is identified as a subset of frames of the second buffer. Correlations of extracted features of the control group to extracted features of successive windows of frames of the first buffer are computed, the extracted features being based on spatial information and temporal information of the video stream. A delay between the video stream collected at the first point and the video stream collected at the second point is identified according to a maximum correlation of the correlations.

Real-time latency measurement of video streams

A source quality of a source video and a source content complexity of the source video are identified. Parameter constraints with respect to parameters of an operation are received. The source video quality, source content complexity, and parameter constraints are applied to a deep neural network (DNN) producing DNN outputs. In an example, the DNN outputs are combined using domain knowledge to provide the filter parameters, as predicted, to a filter chain, such that applying the filter chain to the input source video results in an output video achieving the full reference video quality score. In another example, the DNN outputs are combined using domain knowledge to provide the filter parameters, as predicted, to a filter chain, such that applying the filter chain to the input source video results in an output video achieving the full reference video quality score.

Growing community of inventors

Waterloo, Canada

Jiheng Wang