US 12,456,298 B2
Method for determining video coding test sequence, electronic device and computer storage medium
Jiaojiao Zhang, Beijing (CN)
Assigned to BEIJING BAIDU NETCOM SCIENCE TECHNOLOGY CO., LTD., Beijing (CN)
Filed by Beijing Baidu Netcom Science Technology Co., Ltd., Beijing (CN)
Filed on Sep. 2, 2022, as Appl. No. 17/902,521.
Claims priority of application No. 202111319210.3 (CN), filed on Nov. 9, 2021.
Prior Publication US 2022/0415046 A1, Dec. 29, 2022
Int. Cl. G06V 20/40 (2022.01); H04N 19/90 (2014.01)
CPC G06V 20/41 (2022.01) [H04N 19/90 (2014.11)] 18 Claims
OG exemplary drawing
 
1. A method for determining a video coding test sequence, the method comprising:
determining a candidate video set including multiple candidate videos corresponding to a target service requirement;
classifying the candidate videos by content categories to obtain a target distribution of content categories;
clustering the candidate videos by values of a preset coding complexity to obtain multiple video classes;
selecting from each of the video classes respectively a target class-representative video such that an actual distribution of content categories is consistent with the target distribution of content categories, comprising
determining a candidate video having an actual preset coding complexity closest to the cluster center in each of the video classes as an initial class representative video of the corresponding video class,
classifying the initial class representative videos by content categories to obtain an initial distribution of content categories,
in response to the initial distribution of content categories being inconsistent with the target distribution of content categories, determining a target video class causing the initial distribution of content categories to be inconsistent with the target distribution of content categories,
determining a candidate video in the target video class as the target class-representative video, wherein with the determined candidate video in the target video class, a corresponding distribution of content categories is consistent with the target distribution of content categories, comprising
determining an order of selecting candidate videos in the target video class as new class representative videos based on a distance to the cluster center of the target video class,
in response to the new class representative video selected according to the selection order still making a current distribution of content categories inconsistent with the target distribution of content categories, replacing the new class representative video according to the selection order until the corresponding distribution of content categories is consistent with the target distribution of content categories,
determining the new class representative video in the target video class as the target class-representative video, wherein with the new class representative video, the corresponding distribution of content categories is consistent with the target distribution of content categories; and
constructing a target video coding test sequence based on the target class-representative videos.