公开(公告)号：US11734795B2

公开(公告)日：2023-08-22

申请号：US17135263

申请日：2020-12-28

Applicant: INDUSTRIAL TECHNOLOGY RESEARCH INSTITUTE

Inventor： Yan-Jen Su ,  Ruen-Rone Lee ,  Feng-Sheng Lin

IPC: G06K9/00 ,  G06T3/40 ,  G06T7/55 ,  G06F3/01

CPC classification number: G06T3/4038 ,  G06F3/011 ,  G06F3/017 ,  G06T7/55 ,  G06T2207/10024 ,  G06T2207/20084

Abstract: A processing device and method for a panoramic image includes converting the panoramic image to a cubemap format; under the cubemap format, calculating a depth information image of the panoramic image, and performing a seamless processing on the depth information image. The format of the depth information image is converted to the original format from the cubemap format to obtain the depth information image corresponding to the panoramic image.

公开(公告)号：US11610391B2

公开(公告)日：2023-03-21

申请号：US16730145

申请日：2019-12-30

Applicant: INDUSTRIAL TECHNOLOGY RESEARCH INSTITUTE

Inventor： Feng-Sheng Lin ,  Chu-Fang Wang ,  Ruen-Rone Lee

IPC: G06V10/75 ,  G06V20/10 ,  G06V20/40

Abstract: A cross-domain image comparison method and a cross-domain image comparison system are provided. The cross-domain image comparison method includes the following steps. Two videos in cross-domain are obtained. The videos are generated by different types of devices. A plurality of semantic segmentation areas are obtained from one frame of each of the videos. A region of interest pair (ROI pair) is obtained according to moving paths of the semantic segmentation areas in the videos. Two bounding boxes and two central points of the ROI pair are obtained. A similarity between the frames is obtained according to the bounding boxes and the central points.

公开(公告)号：US11979620B2

公开(公告)日：2024-05-07

申请号：US17554623

申请日：2021-12-17

Applicant: Industrial Technology Research Institute

Inventor： Sheng-Han Wu ,  Feng-Sheng Lin ,  Ruen-Rone Lee

IPC: G06T7/00 ,  G06T7/194 ,  G06V20/40 ,  H04N5/222 ,  H04N5/272 ,  H04N21/234

CPC classification number: H04N21/23424 ,  G06T7/194 ,  G06V20/42 ,  H04N5/2224 ,  H04N5/2723 ,  G06T2207/20081 ,  G06T2207/20084

Abstract: A system for automatically placing virtual advertisements in sports videos, which includes a shot detection module, background extraction module, a calibration module and an asset placement module. The shot detection module detects the target shot of a sports video via a first trained model. The background extraction module performs a background extraction to extract the background of the first frame of the target shot to obtain a first background mask. The calibration module performs a camera calibration to detect a first transformation relation, between the first frame and a sport field template, via a second trained model. The asset placement module transforms an advertisement asset according to the first transformation relation to obtain a first transformed asset, and execute an asset placement to place the first transformed asset onto the first frame according to the first background mask to obtain a first image frame with placed advertisement.

公开(公告)号：US12020481B2

公开(公告)日：2024-06-25

申请号：US17506805

申请日：2021-10-21

Applicant: INDUSTRIAL TECHNOLOGY RESEARCH INSTITUTE

Inventor： Feng-Sheng Lin ,  Sheng-Han Wu ,  Ruen-Rone Lee

IPC: G06V20/40 ,  G06T3/00 ,  G06T3/18 ,  G06V10/94

CPC classification number: G06V20/42 ,  G06T3/18 ,  G06V10/95 ,  G06V20/46

Abstract: A sports event video processing method is provided. The processing method includes: receiving a sports event input video; performing SOI detection on the sports event input video to obtain at least one SOI; performing logo detection and extraction on the at least one SOI to detect at least one logo; performing pixel-level rearrangement on the at least one detected logo; and outputting a sports event output video having completed pixel-level rearrangement.