-
Patent Agency Ranking
公开(公告)号：US20180359414A1
公开(公告)日：2018-12-13
申请号：US15619702
申请日：2017-06-12
Applicant: Adobe Systems Incorporated
Inventor： Oliver Wang , Chengzhou Tang
CPC classification number: H04N5/23238 , G03B37/04 , G06T2200/32 , H04N5/3415 , H04N13/243
Abstract: A technique for modifying digital video includes receiving a plurality of digital video frames recorded by a camera. Each frame has a spherical field of view and a viewing angle associated therewith, where the viewing angle is with respect to a fixed reference frame. A motion of the camera relative to the fixed reference frame is calculated across at least some of the digital video frames. The viewing angle associated with each digital video frame is reoriented during post-processing of the digital video frames based at least in part on the calculated motion of the camera and at least one constraint to produce a digitally modified video such the viewing angle associated with at least one of the reoriented digital video frames is different than the viewing angle associated with the same digital video frame before reorientation.
-
公开(公告)号：US10055898B1
公开(公告)日：2018-08-21
申请号：US15439767
申请日：2017-02-22
Applicant: Adobe Systems Incorporated
Inventor： Jue Wang , Zhaopeng Cui , Oliver Wang
Abstract: Multi-video registration for video synthesis is described. In example implementations, at least one computing device synthesizes multiple videos to create merged images using an automated mechanism to register the multiple videos. The computing device obtains multiple videos with each video including a sequence of multiple frames. Using multiple camera poses determined in a three-dimensional scene reconstruction, respective frames of respective ones of the multiple videos are linked to produce linked frames. The computing device aligns the linked frames to produce aligned frames using point guidance that is based on the multiple spatial points identified in the 3D scene reconstruction. For example, pixels in each of the linked frames that correspond to a same spatial point of the three-dimensional scene reconstruction can be used to align the linked frames at a pixel level. Based on the aligned frames, the computing device creates at least one merged image to synthesize the multiple videos.
-
公开(公告)号：US20180121767A1
公开(公告)日：2018-05-03
申请号：US15341875
申请日：2016-11-02
Applicant: ADOBE SYSTEMS INCORPORATED
Inventor： Oliver Wang , Jue Wang , Shuochen Su
CPC classification number: G06N3/08 , G06K9/6256 , G06K9/6262 , G06N3/04 , G06T5/003 , G06T5/50 , G06T11/60 , G06T2207/10016 , G06T2207/20081 , G06T2207/20084 , G06T2207/20201 , G06T2207/20221
Abstract: Methods and systems are provided for deblurring images. A neural network is trained where the training includes selecting a central training image from a sequence of blurred images. An earlier training image and a later training image are selected based on the earlier training image preceding the central training image in the sequence and the later training image following the central training image in the sequence and based on proximity of the images to the central training image in the sequence. A training output image is generated by the neural network from the central training image, the earlier training image, and the later training image. Similarity is evaluated between the training output image and a reference image. The neural network is modified based on the evaluated similarity. The trained neural network is used to generate a deblurred output image from a blurry input image.
-
公开(公告)号：US20180232906A1
公开(公告)日：2018-08-16
申请号：US15430659
申请日：2017-02-13
Applicant: Adobe Systems Incorporated
Inventor： Vladimir Kim , Oliver Wang , Minhyuk Sung , Mehmet Ersin Yumer
CPC classification number: G06T7/75 , G06T5/002 , G06T7/246 , G06T7/251 , G06T2200/04 , G06T2207/10012 , G06T2207/10021 , G06T2207/10028 , G06T2207/20024 , G06T2207/30241 , G06T2207/30244
Abstract: Disclosed are techniques for more accurately estimating the pose of a camera used to capture a three-dimensional scene. Accuracy is enhanced by leveraging three-dimensional object priors extracted from a large-scale three-dimensional shape database. This allows existing feature matching techniques to be augmented by generic three-dimensional object priors, thereby providing robust information about object orientations across multiple images or frames. More specifically, the three-dimensional object priors provide a unit that is easier and more reliably tracked between images than a single feature point. By adding object pose estimates across images, drift is reduced and the resulting visual odometry techniques are more robust and accurate. This eliminates the need for three-dimensional object templates that are specifically generated for the imaged object, training data obtained for a specific environment, and other tedious preprocessing steps. Entire object classes identified in a three-dimensional shape database can be used to train an object detector.
-
公开(公告)号：US20180240279A1
公开(公告)日：2018-08-23
申请号：US15439767
申请日：2017-02-22
Applicant: Adobe Systems Incorporated
Inventor： Jue Wang , Zhaopeng Cui , Oliver Wang
CPC classification number: G06T19/20 , G06T7/262 , G06T7/37 , G06T17/205 , G06T2207/10016 , G06T2207/20048 , G06T2207/30244 , G06T2215/16 , G06T2219/2004 , G06T2219/2016
Abstract: Multi-video registration for video synthesis is described. In example implementations, at least one computing device synthesizes multiple videos to create merged images using an automated mechanism to register the multiple videos. The computing device obtains multiple videos with each video including a sequence of multiple frames. Using multiple camera poses determined in a three-dimensional scene reconstruction, respective frames of respective ones of the multiple videos are linked to produce linked frames. The computing device aligns the linked frames to produce aligned frames using point guidance that is based on the multiple spatial points identified in the 3D scene reconstruction. For example, pixels in each of the linked frames that correspond to a same spatial point of the three-dimensional scene reconstruction can be used to align the linked frames at a pixel level. Based on the aligned frames, the computing device creates at least one merged image to synthesize the multiple videos.
-
-
-
-
Search Results
5
records
(took 0.015 seconds)
