公开(公告)号：US11870994B2

公开(公告)日：2024-01-09

申请号：US16610467

申请日：2019-03-06

Applicant: PEKING UNIVERSITY SHENZHEN GRADUATE SCHOOL

Inventor： Ronggang Wang ,  Zhenyu Wang ,  Wen Gao

IPC: H04N19/124 ,  H04N19/103 ,  H04N19/136 ,  H04N19/176 ,  H04N19/196

CPC classification number: H04N19/124 ,  H04N19/103 ,  H04N19/136 ,  H04N19/176 ,  H04N19/196

Abstract: A method, system, device and computer-readable storage medium for inverse quantization. The method comprises: determining an initial weighted inverse quantization matrix, wherein, the initial weighted inverse quantization matrix is the same as the quantized block in size; setting some matrix elements in the initial weighted inverse quantization matrix to zero to obtain a weighted inverse quantization matrix, wherein, determining the matrix elements that need to be zeroed according to the size of the quantized block; weighted inverse quantizing the quantized coefficients in the quantized block to generate corresponding inverse transform coefficients, wherein, the value of the matrix element corresponding to the position of the quantized coefficient in the weighted inverse quantization matrix is used as a weight coefficient of the weighted inverse quantization.

公开(公告)号：US11381812B2

公开(公告)日：2022-07-05

申请号：US16755350

申请日：2018-09-25

Applicant: PEKING UNIVERSITY SHENZHEN GRADUATE SCHOOL

Inventor： Ronggang Wang ,  Kui Fan ,  Ge Li ,  Wen Gao

IPC: H04N19/11 ,  H04N19/117 ,  H04N19/147 ,  H04N19/159 ,  H04N19/593 ,  H04N19/82

Abstract: Disclosed is a boundary filtering method for intra prediction, relating to the video encoding technology filed. Whether boundary filtering is performed on an intra prediction block or not is adaptively selected by means of a rate distortion optimization decision; during filtering, a filter coefficient exponentially attenuated relative to distance to boundary is adopted to perform filtering on the first N rows or the first N columns of the intra prediction block by means of an intra prediction block filter, and different filtering strengths are used according to different sizes of the prediction blocks. Therefore, the boundary distortion problem of intra prediction block is solved, the intra prediction precision is improved, and the encoding efficiency of intra prediction block is increased; and the practicability and the robustness of the boundary filtering technology are improved.

公开(公告)号：US11379711B2

公开(公告)日：2022-07-05

申请号：US16414783

申请日：2017-08-16

Applicant: Peking University Shenzhen Graduate School

Inventor： Wenmin Wang ,  Zhihao Li ,  Ronggang Wang ,  Ge Li ,  Shengfu Dong ,  Zhenyu Wang ,  Ying Li ,  Hui Zhao ,  Wen Gao

IPC: G06K9/62 ,  G06N3/04 ,  G06F17/15 ,  G06F17/18 ,  G06N3/08

Abstract: A video action detection method based on a convolutional neural network (CNN) is disclosed in the field of computer vision recognition technologies. A temporal-spatial pyramid pooling layer is added to a network structure, which eliminates limitations on input by a network, speeds up training and detection, and improves performance of video action classification and time location. The disclosed convolutional neural network includes a convolutional layer, a common pooling layer, a temporal-spatial pyramid pooling layer and a full connection layer. The outputs of the convolutional neural network include a category classification output layer and a time localization calculation result output layer. The disclosed method does not require down-sampling to obtain video clips of different durations, but instead utilizes direct input of the whole video at once, improving efficiency. Moreover, the network is trained by using video clips of the same frequency without increasing differences within a category, thus reducing the learning burden of the network, achieving faster model convergence and better detection.

公开(公告)号：US11347979B2

公开(公告)日：2022-05-31

申请号：US16079660

申请日：2016-03-10

Applicant: Peking University Shenzhen Graduate School

Inventor： Wenmin Wang ,  Ruonan Zhang ,  Ronggang Wang ,  Ge Li ,  Shengfu Dong ,  Zhenyu Wang ,  Ying Li ,  Hui Zhao ,  Wen Gao

IPC: G06N7/00 ,  G06K9/62 ,  G06F17/18 ,  G06N5/00 ,  G06N20/00 ,  G06V10/75

Abstract: A method and a device for MCMC framework-based sub-hypergraph matching are provided. Matching of object features is performed by constructing sub-hypergraphs. In a large number of actual images and videos, objects vary constantly, and contain various noise points as well as other interference factors, which makes image object matching and searching very difficult. Perform object feature matching by representing the appearance and positions of objects by sub-hypergraphs allows for faster and more accurate image matching. Furthermore, a sub-hypergraph has several advantages over a graph or a hypergraph: on one hand, a sub-hypergraph has more geometric information (e.g. angle transformation, rotation, scale, etc.) than a graph, and has a lower degree of difficulty and better extensibility than a hypergraph. On the other hand, the disclosed method and device have stronger capabilities to resist interference and good robustness, and are adaptable to more complex settings, especially with outliers.

公开(公告)号：US11301953B2

公开(公告)日：2022-04-12

申请号：US16650141

申请日：2018-05-29

Applicant: PEKING UNIVERSITY SHENZHEN GRADUATE SCHOOL

Inventor： Ronggang Wang ,  Yueming Wang ,  Zhenyu Wang ,  Wen Gao

IPC: G06T3/00 ,  G06T3/40

Abstract: Disclosed are a panoramic video asymmetrical mapping method and a corresponding inverse mapping method that include mapping a spherical surface corresponding to a panoramic image or video A onto a two-dimensional image or video B, projecting the spherical surface onto an isosceles quadrangular pyramid with a square bottom plane, and further projecting the isosceles quadrangular pyramid onto a planar surface, using isometric projection on a main viewpoint region in the projection and using a relatively high sampling density to ensure that the video quality of the region of the main viewpoint is high, while using a relatively low sample density for non-main viewpoint regions so as to reduce bit rate. The panoramic video asymmetrical inverse mapping technique provides a method for mapping from a planar surface to a spherical surface, and a planar surface video may be mapped back to a spherical surface for rendering and viewing.

公开(公告)号：US11216985B2

公开(公告)日：2022-01-04

申请号：US17045894

申请日：2018-05-15

Applicant: Peking University Shenzhen Graduate School

Inventor： Ge Li ,  Qi Zhang ,  Yiting Shao ,  Wen Gao

IPC: G06T9/00 ,  G06T3/40 ,  G06T9/40

Abstract: Disclosed in the present invention is a point cloud attribution compression method based on deleting 0 elements in a quantisation matrix, including optimizing a traversal sequence for a quantisation matrix and deleting the 0 elements at the end of the data stream. The present invention may use seven types of traversal sequences at the encoding end of the point cloud attribute compression, such that the distribution of the 0 elements in the data stream may be more concentrated at the end thereof. The 0 elements at the end of the data stream may be deleted, removing redundant information and reducing the quantity of data to be entropy encoded. At the decoding end, the point cloud geometric information may be incorporated to supplement the deleted 0 elements and the quantisation matrix may be restored according to the traversal sequence, thereby improving compression performance without introducing new errors.

公开(公告)号：US11004240B2

公开(公告)日：2021-05-11

申请号：US16626907

申请日：2018-05-15

Applicant: Peking University Shenzhen Graduate School

Inventor： Ge Li ,  Yi Ting Shao ,  Qi Zhang ,  Rong Gang Wang ,  Tie Jun Huang ,  Wen Gao

IPC: G06T9/40 ,  G06T7/90 ,  G06T7/60

Abstract: Disclosed is a hierarchical division-based point cloud attribute compression method. For point cloud attribute information, a new hierarchical division based coding scheme is proposed, wherein a frame of point cloud is adaptively divided into a “stripe-macroblock-block” hierarchical structure according to the spatial position and color distribution of the point cloud, and stripes are coded independently from one another, increasing the coding efficiency, enhancing the fault tolerance of a system and improving the performance of point cloud attribute compression. The method comprises: (1) inputting a point cloud; (2) division of a k-dimension (KD) tree structure of the point cloud; (3) continuity analysis of point cloud attribute information; (4) stripe division of the point cloud; (5) division of macroblocks and coding blocks of the point cloud; and (6) intra-frame prediction, transformation, quantification and entropy coding based on a block structure.

公开(公告)号：US10719664B2

公开(公告)日：2020-07-21

申请号：US16314673

申请日：2016-12-01

Applicant: Peking University Shenzhen Graduate School

Inventor： Wenmin Wang ,  Liang Han ,  Mengdi Fan ,  Ronggang Wang ,  Ge Li ,  Shengfu Dong ,  Zhenyu Wang ,  Ying Li ,  Hui Zhao ,  Wen Gao

IPC: G06F16/00 ,  G06T11/60 ,  G06F40/30 ,  G06F40/216 ,  G06F40/284 ,  G06N20/00 ,  G06K9/00 ,  G06K9/62 ,  G06N3/08 ,  G06N7/00

Abstract: A cross-media search method using a VGG convolutional neural network (VGG net) to extract image features. The 4096-dimensional feature of a seventh fully-connected layer (fc7) in the VGG net, after processing by a ReLU activation function, serves as image features. A Fisher Vector based on Word2vec is utilized to extract text features. Semantic matching is performed on heterogeneous images and the text features by means of logistic regression. A correlation between the two heterogeneous features, which are images and text, is found by means of semantic matching based on logistic regression, and thus cross-media search is achieved. The feature extraction method can effectively indicate deep semantics of image and text, improve cross-media search accuracy, and thus greatly improve the cross-media search effect.

公开(公告)号：US10339409B2

公开(公告)日：2019-07-02

申请号：US15575897

申请日：2015-06-18

Applicant: Peking University Shenzhen Graduate School

Inventor： Wenmin Wang ,  Mingmin Zhen ,  Ronggang Wang ,  Ge Li ,  Shengfu Dong ,  Zhenyu Wang ,  Ying Li ,  Wen Gao

IPC: G06K9/46 ,  G06T7/64 ,  G06K9/42 ,  G06T5/40 ,  G06K9/00

Abstract: A method and a device for extracting local features of a 3D point cloud are disclosed. Angle information and the concavo-convex information about a feature point to be extracted and a point of an adjacent body element are calculated based on a local reference system corresponding to the points of each body element. The feature relation between the two points can be calculated accurately. The property of invariance in translation and rotation is possessed. Since concavo-convex information about a local point cloud is contained during extraction, the inaccurate extraction caused by ignoring concavo-convex ambiguity in previous 3D local feature description is resolved. During normalization processing, exponential normalization processing and second-normal-form normalization are adopted, which solves the problem of inaccurate similarity calculation caused by a circumstance that a few elements in a vector are too large or too small during feature extraction, thus improving accuracy of extracted three-dimensional local features.

公开(公告)号：US10298950B2

公开(公告)日：2019-05-21

申请号：US15006147

申请日：2016-01-26

Applicant: PEKING UNIVERSITY SHENZHEN GRADUATE SCHOOL

Inventor： Ronggang Wang ,  Lei Chen ,  Zhenyu Wang ,  Siwei Ma ,  Wen Gao ,  Tiejun Huang ,  Wenmin Wang ,  Shengfu Dong

IPC: H04N19/51 ,  H04N19/52 ,  H04N19/55 ,  H04N19/56 ,  H04N19/91 ,  H04N19/176 ,  H04N19/513 ,  H04N19/533 ,  H04N19/553 ,  H04N19/557 ,  H04N19/157 ,  H04N19/159

Abstract: A P frame-based multi-hypothesis motion compensation method includes: taking an encoded image block adjacent to a current image block as a reference image block and obtaining a first motion vector of the current image block by using a motion vector of the reference image block, the first motion vector pointing to a first prediction block; taking the first motion vector as a reference value and performing joint motion estimation on the current image block to obtain a second motion vector of the current image block, the second motion vector pointing to a second prediction block; and performing weighted averaging on the first prediction block and the second prediction block to obtain a final prediction block of the current image block. The method increases the accuracy of the obtained prediction block of the current image block without increasing the code rate.