知搜-全球专利检索

  1. Login
  2. Sign Up

Search Results

9 records (took 0.013 seconds)

    IMAGE CLASSIFICATION METHOD BASED ON RELIABLE WEIGHTED OPTIMAL TRANSPORT (RWOT)
    1.
    发明申请
    IMAGE CLASSIFICATION METHOD BASED ON RELIABLE WEIGHTED OPTIMAL TRANSPORT (RWOT) 有权

    公开(公告)号：US20210390355A1

    公开(公告)日：2021-12-16

    申请号：US17347546

    申请日：2021-06-14

    Applicant: Zhejiang University

    Inventor： Renjun Xu ,  Weiming Liu ,  Jiuming Lin ,  Xinyue Qian ,  Xiaoyue Hu ,  Yin Zhao ,  Jingcheng He ,  Zihang Zhu ,  Xu He ,  Chengbo Sun ,  Xiang Zhou

    IPC: G06K9/62 ,  G06N3/08 ,  G06N3/04

    Abstract: An image classification method based on reliable weighted optimal transport (RWOT) includes: preprocessing data in a source domain, so that a deep neural network fits a sample image in the source domain to obtain a sample label; performing image labeling to add a pseudo label to a data sample in a target domain; performing node pairing to pair associated images in the source domain and the target domain; and performing automatic analysis by using a feature extractor and an adaptive discriminator, to perform image classification. The present disclosure proposes a subspace reliability method for dynamically measuring a difference between the source domain and the target domain based on spatial prototypical information and an intra-domain structure. This method can be used as a preprocessing step of an existing domain adaptation technology, and greatly improves efficiency.

    Method and device for generating predicted pictures
    4.
    发明授权
    Method and device for generating predicted pictures 有权

    公开(公告)号：US10154282B2

    公开(公告)日：2018-12-11

    申请号：US14441445

    申请日：2013-09-06

    Applicant: ZHEJIANG UNIVERSITY ,  ZTE Corporation

    Inventor： Lu Yu ,  Yichen Zhang ,  Yin Zhao ,  Yingjie Hong ,  Ming Li

    IPC: H04N19/583 ,  H04N19/597 ,  H04N19/105 ,  H04N19/167 ,  H04N19/176 ,  H04N19/52 ,  H04N19/157

    Abstract: Disclosed are a method and device for generating a predicted picture, the method comprising: determining a reference rectangular block of pixels according to parameter information which includes a location of a target rectangular block of pixels and/or depth information of a reference view; mapping the reference rectangular block of pixels to a target view according to the depth information of the reference view to obtain a projection rectangular block of pixels; and acquiring a predicted picture block from the projection rectangular block of pixels. The technical problem of relatively large dependence among the data brought by simultaneously employing the depth picture of the target view and the depth picture of the reference view in the process of generating the predicted picture in the prior art is solved, and the technical effects of reducing the dependence on the data and improving the encoding and decoding efficiency is achieved.

    Method and device for generating predicted pictures
    6.
    发明授权
    Method and device for generating predicted pictures 有权

    公开(公告)号：US10708619B2

    公开(公告)日：2020-07-07

    申请号：US16168289

    申请日：2018-10-23

    Applicant: SHANGHAI ZHONGXING SOFTWARE COMPANY LIMITED ,  ZHEJIANG UNIVERSITY

    Inventor： Lu Yu ,  Yichen Zhang ,  Yin Zhao ,  Yingjie Hong ,  Ming Li

    IPC: H04N19/583 ,  H04N19/176 ,  H04N19/597 ,  H04N19/105 ,  H04N19/167 ,  H04N19/52 ,  H04N19/157

    Abstract: Disclosed are a method and device for generating a predicted picture, the method comprising: determining a reference rectangular block of pixels by shifting or warping parameter information which includes a location of a target rectangular block of pixels, or the parameter information comprises a location of a target rectangular block of pixels and depth information of a reference view; mapping the reference rectangular block of pixels to a target view according to the depth information of the reference view to obtain a projection rectangular block of pixels; and acquiring a predicted picture block from the projection rectangular block of pixels. The technical problem of relatively large dependence among the data brought by simultaneously employing the depth picture of the target view and the depth picture of the reference view in the process of generating the predicted picture in the prior art is solved, and the technical effects of reducing the dependence on the data and improving the encoding and decoding efficiency is achieved.

    Method and device for generating stereoscopic video pair
    7.
    发明授权
    Method and device for generating stereoscopic video pair 有权

    公开(公告)号：US09888222B2

    公开(公告)日：2018-02-06

    申请号：US14397123

    申请日：2013-04-24

    Applicant: ZHEJIANG UNIVERSITY

    Inventor： Lu Yu ,  Yin Zhao

    IPC: H04N13/00 ,  G06T3/40

    CPC classification number: H04N13/111 ,  G06T3/40 ,  H04N13/122 ,  H04N2213/003 ,  H04N2213/007

    Abstract: The present invention discloses a method to generate a stereoscopic video pair which can be used in the field of multimedia communication. A stereoscopic video pair can be obtained after processing a stereoscopic video sequence by using display auxiliary information. The stereoscopic video pair is displayed on a stereoscopic display D1. The display auxiliary information comprises camera viewpoint position information, virtual viewpoint position information and display scaling factor S1. The camera viewpoint position information indicates the position of a camera viewpoint C of a three-dimensional video sequence. The virtual viewpoint position information indicates the position of a virtual viewpoint P1. The display scaling factor S1 indicates the ratio of horizontal resolution Res1 to horizontal width W1 of the stereoscopic display D1, i.e., S1=Res1/W1, where the horizontal width is the approximation of actual display screen width. The present invention also discloses the corresponding stereoscopic video pair generation apparatus. The present invention can improve visual experience of stereoscopic viewing.

    METHOD AND DEVICE FOR GENERATING PREDICTED PICTURES
    8.
    发明申请
    METHOD AND DEVICE FOR GENERATING PREDICTED PICTURES 审中-公开
    用于产生预测图像的方法和装置

    公开(公告)号：US20150296219A1

    公开(公告)日：2015-10-15

    申请号：US14441445

    申请日：2013-09-06

    Applicant: Zhejiang University ,  ZTE Corporation

    Inventor： Lu Yu ,  Yichen Zhang ,  Yin Zhao ,  Yingjie Hong ,  Ming Li

    IPC: H04N19/583 ,  H04N19/52

    CPC classification number: H04N19/583 ,  H04N19/105 ,  H04N19/157 ,  H04N19/167 ,  H04N19/176 ,  H04N19/52 ,  H04N19/597

    Abstract: Disclosed are a method and device for generating a predicted picture, the method comprising: determining a reference rectangular block of pixels according to parameter information which includes a location of a target rectangular block of pixels and/or depth information of a reference view; mapping the reference rectangular block of pixels to a target view according to the depth information of the reference view to obtain a projection rectangular block of pixels; and acquiring a predicted picture block from the projection rectangular block of pixels. The technical problem of relatively large dependence among the data brought by simultaneously employing the depth picture of the target view and the depth picture of the reference view in the process of generating the predicted picture in the prior art is solved, and the technical effects of reducing the dependence on the data and improving the encoding and decoding efficiency is achieved.

    Abstract translation: 公开了一种用于生成预测图像的方法和装置，所述方法包括：根据包括目标矩形块像素的位置和/或参考视图的深度信息的参数信息来确定参考矩形块像素; 根据参考视图的深度信息将参考矩形块的像素映射到目标视图，以获得投影矩形的像素块; 以及从所述投影矩形像素块获取预测图像块。 解决了在现有技术中生成预测图像的过程中同时采用目标视图的深度图像和参考视图的深度图像所带来的数据之间相对较大相关性的技术问题，并且减少了技术效果 实现对数据的依赖，提高编码和解码效率。

    METHOD AND DEVICE FOR DECODING AND ENCODING SUPPLEMENTAL AUXILIARY INFORMATION OF THREE-DIMENSIONAL VIDEO SEQUENCE
    9.
    发明申请
    METHOD AND DEVICE FOR DECODING AND ENCODING SUPPLEMENTAL AUXILIARY INFORMATION OF THREE-DIMENSIONAL VIDEO SEQUENCE 审中-公开
    用于解码和编码三维视频序列辅助辅助信息的方法和装置

    公开(公告)号：US20150117547A1

    公开(公告)日：2015-04-30

    申请号：US14397158

    申请日：2013-04-24

    Applicant: ZTE CORPORATION ,  Zhejiang University

    Inventor： Yin Zhao ,  Lu Yu ,  Yingjie Hong ,  Ming Li

    IPC: H04N13/00

    CPC classification number: H04N13/161 ,  H04N13/111 ,  H04N13/172 ,  H04N13/178 ,  H04N21/2353 ,  H04N21/2362 ,  H04N21/4345 ,  H04N21/816

    Abstract: Provided are a method and device for decoding and encoding supplemental auxiliary information of a three-dimensional video sequence. The method includes: obtaining supplemental auxiliary information for constructing a stereo-pair from a three-dimensional video sequence bitstream, the supplemental auxiliary information being used for indicating that the stereo-pair is constructed from a reconstructed three-dimensional video sequence, and the reconstructed three-dimensional video sequence being obtained by decoding the three-dimensional video sequence bitstream. The present invention solves the technical problem in the prior art that degraded displaying quality of a constructed stereo-pair appears because of the lack of supplemental auxiliary information in a three-dimensional video sequence bitstream, and achieves the technical effect of improving the display quality of a constructed stereoscopic video.

    Abstract translation: 提供了一种用于对三维视频序列的补充辅助信息进行解码和编码的方法和装置。 该方法包括：从三维视频序列比特流中获取用于构建立体对的辅助辅助信息，所述补充辅助信息用于指示立体声是由重建的三维视频序列构成的， 通过解码三维视频序列位流获得三维视频序列。 本发明解决了现有技术中由于在三维视频序列比特流中缺少补充辅助信息而导致的立体对的劣化显示质量出现的技术问题，并且达到提高显示质量的技术效果 构建的立体视频。

Patent Agency Ranking