CPC H04N 19/136 (2014.11) [G06N 20/00 (2019.01); H04N 19/103 (2014.11); H04N 19/179 (2014.11); H04N 21/23439 (2013.01); H04N 21/8456 (2013.01)] | 30 Claims |
1. A method, comprising:
in a training system:
from a plurality of videos that are decoded from coded video data, extracting features of the decoded videos according to their content,
applying a machine learning algorithm to the extracted features of the decoded videos, subjective review scores of one or more viewers of the decoded videos, and coding parameters used to encode the decoded videos to generate a model of coding parameters, and
building, based on the model of coding parameters, a database of coding parameters and associated extracted features of the videos associated with high subjective review scores; and
exporting the database from the training system to a consumer device for use in comparing features extracted from a video captured by the consumer device to the extracted features in a database to determine coding parameters.
|