FIELD: physics.
SUBSTANCE: group of inventions relates to computer vision. Disclosed is a method of correcting depth data for a plurality of views in a video sequence comprising one or more frames, which comprises a step, on which frame-by-frame depth maps are created in the form of bilinear splines for a subset of frames from said one or more frames. For the remaining frames in one or more frames, frame-by-frame depth maps are initialized by reprojection of metric depth maps of this subset of frames. For all frames from one or more frames performing iterative optimization of frame-by-frame depth maps, wherein all scaled depth maps are initialized in units before correction; and performing iterative optimization of depth scaling maps for said subset of frames by means of a gradient-based optimization algorithm. Device and machine-readable medium for implementing said method are also provided.
EFFECT: obtaining consistent metric depth maps based on absolute distances to all objects in a medium displayed in a sequence of video frames.
13 cl, 4 dwg, 2 tbl
Title | Year | Author | Number |
---|---|---|---|
VISUALIZATION OF RECONSTRUCTION OF 3D SCENE USING SEMANTIC REGULARIZATION OF NORMALS TSDF WHEN TRAINING NEURAL NETWORK | 2023 |
|
RU2825722C1 |
METHOD AND SYSTEM FOR REFINING THE CAMERA POSITION TAKING INTO ACCOUNT THE ROOM PLAN | 2022 |
|
RU2794441C1 |
SYSTEMS AND METHODS OF ASSESSING THE VIABILITY OF EMBRYOS | 2018 |
|
RU2800079C2 |
METHOD FOR ESTIMATING THE DEPTH OF A SCENE BASED ON AN IMAGE AND COMPUTING APPARATUS FOR IMPLEMENTATION THEREOF | 2020 |
|
RU2761768C1 |
SYSTEM AND METHOD FOR VIDEO TEMPORARY COMPLEMENT | 2014 |
|
RU2560086C1 |
METHOD AND ELECTRONIC DEVICE FOR DETECTING THREE-DIMENSIONAL OBJECTS USING NEURAL NETWORKS | 2021 |
|
RU2776814C1 |
METHOD FOR CREATION OF MULTILAYERED SCENE REPRESENTATION AND COMPUTER DEVICE FOR ITS IMPLEMENTATION | 2021 |
|
RU2787928C1 |
METADATA FOR DEPTH FILTRATION | 2013 |
|
RU2639686C2 |
MODELLING PEOPLE'S CLOTHING BASED ON MULTIPLE POINTS | 2021 |
|
RU2776825C1 |
METHOD OF TRAINING A CONVOLUTIONAL NEURAL NETWORK FOR IMAGE RECONSTRUCTION AND A SYSTEM FOR FORMING AN IMAGE DEPTH MAP (VERSIONS) | 2018 |
|
RU2698402C1 |
Authors
Dates
2024-09-26—Published
2023-07-24—Filed