CN108429889A - A kind of 1,000,000,000 pixel video generation method of EO-1 hyperion - Google Patents

A kind of 1,000,000,000 pixel video generation method of EO-1 hyperion Download PDF

Info

Publication number
CN108429889A
CN108429889A CN201810183539.3A CN201810183539A CN108429889A CN 108429889 A CN108429889 A CN 108429889A CN 201810183539 A CN201810183539 A CN 201810183539A CN 108429889 A CN108429889 A CN 108429889A
Authority
CN
China
Prior art keywords
image
view image
local multi
block
local
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN201810183539.3A
Other languages
Chinese (zh)
Inventor
方璐
刘烨斌
王好谦
戴琼海
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shenzhen Weilai Media Technology Research Institute
Shenzhen Graduate School Tsinghua University
Original Assignee
Shenzhen Weilai Media Technology Research Institute
Shenzhen Graduate School Tsinghua University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shenzhen Weilai Media Technology Research Institute, Shenzhen Graduate School Tsinghua University filed Critical Shenzhen Weilai Media Technology Research Institute
Priority to CN201810183539.3A priority Critical patent/CN108429889A/en
Publication of CN108429889A publication Critical patent/CN108429889A/en
Pending legal-status Critical Current

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
    • H04N23/90—Arrangement of cameras or camera modules, e.g. multiple cameras in TV studios or sports stadiums
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/90—Determination of colour characteristics
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N5/00—Details of television systems
    • H04N5/222—Studio circuitry; Studio devices; Studio equipment
    • H04N5/262—Studio circuits, e.g. for mixing, switching-over, change of character of image, other special effects ; Cameras specially adapted for the electronic generation of special effects
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N5/00—Details of television systems
    • H04N5/222—Studio circuitry; Studio devices; Studio equipment
    • H04N5/262—Studio circuits, e.g. for mixing, switching-over, change of character of image, other special effects ; Cameras specially adapted for the electronic generation of special effects
    • H04N5/2628—Alteration of picture size, shape, position or orientation, e.g. zooming, rotation, rolling, perspective, translation

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Image Processing (AREA)

Abstract

A kind of 1,000,000,000 pixel video generation method of EO-1 hyperion, including：Image Acquisition；Pre-processing image data：By the size of the magazine each image in local visual angle according to fr/flIt reduces, frFor the focusing length of local visual angle camera, flFor the focusing length of reference camera；Global change's iteration：Find each local multi-view image and its feature consistency with robustness between corresponding reference image block in reference image；Transformation iteration based on mesh：Single induction method based on mesh transformation improves feature consistency；Color calibration and overlap processing：It assigns the Color Style of any reference block to corresponding local multi-view image, the Color Style of reference is transferred in local multi-view image, transfer method is according to affine color mapping model；Using Graphcut algorithms to replace the local multi-view image being overlapped on reference block on reference block.Using the method for the present invention, 1,000,000,000 pixel video of high-quality can be efficiently and accurately obtained.

Description

A kind of 1,000,000,000 pixel video generation method of EO-1 hyperion
Technical field
The present invention relates to computer visions and field of video processing, are given birth to more particularly to a kind of 1,000,000,000 pixel video of EO-1 hyperion At method.
Background technology
Conventional video systems restrictive premise --- camera resolution needs to match display resolution, as HD videos use HD cameras and display screen, 4K videos use 4K cameras and display screen.However, with 1,000,000,000 pixel techniques and virtual reality technology The information content of development, camera intake is far above the information content that display screen can be shown.These systems can utilize multi-scale image Structure, to reinforce interacting the image stream explored and taken depending on person.
Invention content
It is a primary object of the present invention in view of the deficiencies of the prior art, provide a kind of 1,000,000,000 pixel video of EO-1 hyperion generation Method.
To achieve the above object, the present invention uses following technical scheme：
A kind of 1,000,000,000 pixel video generation method of EO-1 hyperion, the described method comprises the following steps：
A0：Image Acquisition：It is mixed with the multiple dimensioned of reference camera including multiple local visual angle cameras and a global visual angle Close the local multi-view image of camera array acquisition and reference image；
A1：Pre-processing image data：By the size of the magazine each image in local visual angle according to fr/flIt reduces, frFor office The focusing length of portion visual angle camera, flFor the focusing length of reference camera；
A2：Global change's iteration：Find each local multi-view image and its corresponding reference image in reference image Feature consistency with robustness between block；
A3：Transformation iteration based on mesh：Single induction method based on mesh transformation improves feature consistency；
A4：Color calibration and overlap processing：Assign the Color Style of any reference block to corresponding local multi-view image, it will The Color Style of reference is transferred in local multi-view image, and transfer method is according to affine color mapping model；It is answered on reference block With Graphcut algorithms to replace the local multi-view image being overlapped on reference block.
Further：
Pre-processing image data in the step A1 is by each local multi-view video according to fr/flRatio carries out size contracting It puts.
In the step A2, Shandong between each local multi-view image and corresponding reference image block is found out by following steps The feature consistency of stick：
A2.1, phase is normalized to the image block in local multi-view image and using zero-mean between corresponding reference image block Mutual relation number carries out structuring similarity measurement, obtains initial feature consistency；
A2.2, the feature based on successful match calculate global homography matrix, to which local multi-view image is transformed to ginseng According on image block, wherein carrying out global change's iteration twice based on following formula, indicate as follows：
WhereinIndicate that an image block in local multi-view image, π indicate that distant view projection, w are the size of image block, ∈ For search width, IrFor reference picture, IlFor topography, r is reference, and l is local visual angle, and p is to be mapped by three-dimensional homogeneous coordinates The two-dimensional coordinate arrived, prFor the two-dimensional coordinate of reference picture being mapped to by three-dimensional homogeneous coordinates, plIt is topography by three-dimensional The two-dimensional coordinate that homogeneous coordinates are mapped to, pr* it is that right-hand side expression is made to reach maximum prValue, H is homography matrix.
In the step A2.2, in the first iteration, H is initialized as unit matrix, and the size w of image block and search are wide Degree ∈ is both configured to 256, and in second of iteration, w and ∈ are disposed as 128.
In the step A2.2, accelerated by FFT.
In the step A3, error hiding strategy is introduced to the smooth texture region of bulk, including：
A3.1, supplement light stream matching, wherein apply include in second of iteration based on mesh convert as a result, the net The process of hole transformation is the process of feature consistency detection；
Improve each picture between local multi-view image and reference image using the optical flow algorithm of variation based on transformation results The consistency of vegetarian refreshments；
For smooth texture region, using the single pixel consistency after improvement as complementary Optical-flow Feature, these features Significant characteristics or its position including local multi-view image are located at grid vertex；
A3.2, the transformation based on grid use ASAP transformation frameworks, carry out joint transform and stabilisation.
In the step A3.2, stabilized using following minimum energy function：
E (V)=λrEr(V)+λtEt(V)+λsEs(V)
V indicates the position of grid vertex, ErIt indicates that current local multi-view image is forced to transform to approximate reference block image State, EtIndicate temporary stable state, EsIndicate the spatial smoothness deformation of regularization adjacent vertex, λr、λt、λsFor state Parameter, for correcting influence of three states for final energy function.
ErCalculate all feature consistencies distance and,
αplIt is plBilinear interpolation weight；
EtIt is defined as：
Whether B indicates indicator function, for examining some pixel in static background；B(pl)=0 indicates to work as plIt is located at When on mobile object,Refer to 2 dimension coordinates of the matched localview of previous frame, displacements of the s between two frames.
The step A4 includes：
A4.1, the Color Style of any reference block is assigned to corresponding local multi-view image, transfer method is according to affine face Color mapping model；
A4.2, it is directed to existing overlapping region between local multi-view image, Graphcut algorithms is applied on reference block, with Replace the local multi-view image being overlapped on reference block.
Using the method for the present invention, 1,000,000,000 pixel video of high-quality can be efficiently and accurately got.The advantage of the present invention It is mainly reflected in following three aspects：
1, this method avoid locally regarded in the adjustment of accurate camera, tediously long geometry and color calibration and existing method The needs of picture registration in the camera of angle；2, this method allows the movement of local visual angle camera, to by being distributed more for region of interest Multisensor resource is next adaptively, efficiently generates 1,000,000,000 pixel videos；3, this method can be with the final high-resolution of parallel generation Rate video, obtains synthesizing 1,000,000,000 pixel videos in real time and provides possibility for optimization in future.
, can be by using the space reflection relationship of topography to global reference picture using the method for the present invention, it will be complete The multispectral information superpackets of office, which are distinguished, maps to part, and the high-resolution spectroscopy image for completing local camera image generates.Based on this The method of invention can carry out lateral resolution images match and transformation by multiple dimensioned Hybrid camera array, multispectral with center Information MAP is to dividing visual field video.
Description of the drawings
Fig. 1 is a kind of flow chart of embodiment of 1,000,000,000 pixel video generation method of EO-1 hyperion of the present invention.
Specific implementation mode
It elaborates below to embodiments of the present invention.It is emphasized that following the description is only exemplary, The range being not intended to be limiting of the invention and its application.
Referring to Fig. 1, in one embodiment, a kind of 1,000,000,000 pixel video generation method of EO-1 hyperion includes the following steps：
A0：Image Acquisition：It is mixed with the multiple dimensioned of reference camera including multiple local visual angle cameras and a global visual angle Close the local multi-view image of camera array acquisition and reference image；
A1：Pre-processing image data：First by the size of the magazine each image in local visual angle according to fr/flIt reduces, fr For the focusing length of local visual angle camera, flFor the focusing length of reference camera；
A2：Global change's iteration：Find each local multi-view image and its corresponding reference image in reference image Feature consistency with robustness between block；
A3：Transformation iteration based on mesh：Single induction method based on mesh transformation improves feature consistency；
A4：Color calibration and overlap processing：Assign the Color Style of any reference block to corresponding local multi-view image, it will The Color Style of reference is transferred in local multi-view image, and transfer method is according to affine color mapping model；It is answered on reference block With Graphcut algorithms to replace the local multi-view image being overlapped on reference block.This processing mode can also solve to regard due to part Angle camera can move, and overlapping region is inevitably present between camera, the alignment error problem occurred in conversion process.
In a preferred embodiment, the pre-processing image data in the step A1 is by each local multi-view video according to fr/fl Ratio carries out size scaling.
In a preferred embodiment, in the step A2, each local multi-view image and corresponding is found out by following steps The feature consistency of robust between reference image block：
A2.1, phase is normalized to the image block in local multi-view image and using zero-mean between corresponding reference image block Mutual relation number carries out structuring similarity measurement, obtains initial feature consistency；
A2.2, the feature based on successful match calculate global homography matrix, to which local multi-view image is transformed to ginseng According on image block, wherein carrying out global change's iteration twice based on following formula, indicate as follows：
WhereinIndicate that an image block in local multi-view image, π indicate that distant view projection, w are the size of image block, ∈ For search width, IrFor reference picture, IlFor topography, r is reference, and l is local visual angle, and p is to be mapped by three-dimensional homogeneous coordinates The two-dimensional coordinate arrived, prFor the two-dimensional coordinate of reference picture being mapped to by three-dimensional homogeneous coordinates, plIt is topography by three-dimensional The two-dimensional coordinate that homogeneous coordinates are mapped to, pr* it is that right-hand side expression is made to reach maximum prValue, H is homography matrix.
In a preferred embodiment, in the step A2.2, in the first iteration, H is initialized as unit matrix, image block Size w and search width ∈ be both configured to 256, in second of iteration, w and ∈ are disposed as 128.
In a preferred embodiment, in the step A2.2, accelerated by FFT.
In a preferred embodiment, in the step A3, error hiding strategy is introduced to the smooth texture region of bulk, including：
A3.1, supplement light stream matching, wherein apply include in second of iteration based on mesh convert as a result, the net The process of hole transformation is the process of feature consistency detection；
Improve each picture between local multi-view image and reference image using the optical flow algorithm of variation based on transformation results The consistency of vegetarian refreshments；
For smooth texture region, using the single pixel consistency after improvement as complementary Optical-flow Feature, these features Significant characteristics or its position including local multi-view image are located at grid vertex；
A3.2, the transformation based on grid use ASAP transformation frameworks, carry out joint transform and stabilisation.
In the present embodiment, the light stream transformation of transformation and variation based on mesh is the presence to complement one another.In step A2.2, ZNCC principles are applied when application using the mapping mode based on mesh.Processing is regarded based on the transformation of mesh Feel salient region such as structure problem, and light stream matching is excellent for the performance of bulk Non feature regions, light stream transformation is for accidentally Matching strategy has advantage.The transformation based on mesh carries out feature consistency detection in second of iteration.Based on transformation results, Improve the consistency of each pixel between local multi-view image and reference image by the optical flow algorithm of variation.
Transformation of the optical flow algorithm based on mesh provides comparable high quality initialization value.Light stream consistency is selected as mutual The Optical-flow Feature of benefit.These features include that the significant characteristics of local multi-view image or its position are located at grid vertex.
In a more preferred embodiment, it in the step A3.2, is stabilized using following minimum energy function：
E (V)=λrEr(V)+λtEt(V)+λsEs(V)
V indicates the position of grid vertex, ErIt indicates that current local multi-view image is forced to transform to approximate reference block image State, EtIndicate temporary stable state, EsIndicate the spatial smoothness deformation of regularization adjacent vertex, λr、λt、λsFor state Parameter, for correcting influence of three states for final energy function.
ErCalculate all feature consistencies distance and,
αplIt is plBilinear interpolation weight；
EtIt is defined as：
Whether B indicates indicator function, for examining some pixel in static background；B(pl)=0 indicates to work as plIt is located at When on mobile object,Refer to 2 dimension coordinates of the matched localview of previous frame, displacements of the s between two frames.
In a preferred embodiment, the step A4 includes：
A4.1, the Color Style of any reference block is assigned to corresponding local multi-view image, transfer method is according to affine face Color mapping model；
A4.2, it is directed to existing overlapping region between local multi-view image, Graphcut algorithms is applied on reference block, with Replace the local multi-view image being overlapped on reference block.
In an example, include reference camera and 14 local visual cameras using the system of the method for the present invention, Whole cameras are PointGrey FL3-U3-88S2C-C rolling screen door cameras, and spatial resolution is 4000 × 3000, per second 15 Frame.Particularly, local visual angle camera shares 135 millimeters of (fr) focusing lengths to obtain the video of local high resolution, and refers to The focusing length of camera is 16 millimeters (fl) to cover the big visual field in Outdoor Scene.When image data acquiring, each Local visual angle camera can be static or movement.1,000,000,000 pixel videos can be more accurately obtained using the present invention.
The above content is specific/preferred embodiment further description made for the present invention is combined, cannot recognize The specific implementation of the fixed present invention is confined to these explanations.For those of ordinary skill in the art to which the present invention belongs, Without departing from the inventive concept of the premise, some replacements or modification can also be made to the embodiment that these have been described, And these are substituted or variant all shall be regarded as belonging to protection scope of the present invention.

Claims (8)

1. a kind of 1,000,000,000 pixel video generation method of EO-1 hyperion, which is characterized in that the described method comprises the following steps：
A0：Image Acquisition：With the multiple dimensioned mixed phase including multiple local visual angle cameras and the reference camera at a global visual angle Machine array acquisition part multi-view image and reference image；
A1：Pre-processing image data：By the size of the magazine each image in local visual angle according to fr/flIt reduces, frIt is regarded for part The focusing length of angle camera, flFor the focusing length of reference camera；
A2：Global change's iteration：Find each local multi-view image and its in reference image corresponding reference image block it Between with robustness feature consistency；
A3：Transformation iteration based on mesh：Single induction method based on mesh transformation improves feature consistency；
A4：Color calibration and overlap processing：It assigns the Color Style of any reference block to corresponding local multi-view image, will refer to Color Style be transferred in local multi-view image, transfer method is according to affine color mapping model；It is applied on reference block Graphcut algorithms are to replace the local multi-view image being overlapped on reference block.
2. the method as described in claim 1, which is characterized in that the pre-processing image data in the step A1 is by each part Multi-view video is according to fr/flRatio carries out size scaling.
3. the method as described in claim 1, which is characterized in that in the step A2, each part is found out by following steps The feature consistency of robust between multi-view image and corresponding reference image block：
A2.1, mutually pass is normalized to the image block in local multi-view image and using zero-mean between corresponding reference image block Coefficient carries out structuring similarity measurement, obtains initial feature consistency；
A2.2, the feature based on successful match calculate global homography matrix, to transform to local multi-view image with reference to figure As on block, wherein carrying out global change's iteration twice based on following formula, indicating as follows：
WhereinIndicate that an image block in local multi-view image, π indicate that distant view projection, w are the size of image block, ∈ is to search Suo Kuandu, IrFor reference picture, IlFor topography, r indicates that reference, l indicate that local visual angle, p are to be mapped by three-dimensional homogeneous coordinates The two-dimensional coordinate arrived, prFor the two-dimensional coordinate of reference picture being mapped to by three-dimensional homogeneous coordinates, plIt is topography by three-dimensional The two-dimensional coordinate that homogeneous coordinates are mapped to, pr* it is that right-hand side expression is made to reach maximum prValue, H is homography matrix.
4. method as claimed in claim 3, which is characterized in that in the step A2.2, in the first iteration, H initialization For unit matrix, the size w and search width ∈ of image block are both configured to 256, and in second of iteration, w and ∈ are disposed as 128。
5. method as claimed in claim 3, which is characterized in that in the step A2.2, accelerated by FFT.
6. method as claimed in claim 3, which is characterized in that in the step A3, introduced to the smooth texture region of bulk Error hiding strategy, including：
A3.1, supplement light stream matching, wherein apply include in second of iteration based on mesh convert as a result, the mesh change The process changed is the process of feature consistency detection；
Improve each pixel between local multi-view image and reference image using the optical flow algorithm of variation based on transformation results Consistency；
For smooth texture region, using the single pixel consistency after improvement as complementary Optical-flow Feature, these features include The significant characteristics of local multi-view image or its position are located at grid vertex；
A3.2, the transformation based on grid use ASAP transformation frameworks, carry out joint transform and stabilisation.
7. method as claimed in claim 6, which is characterized in that in the step A3.2, use following minimum energy letter Number is stabilized：
E (V)=λrEr(V)+λtEt(V)+λsEs(V)
V indicates the position of grid vertex, ErIndicate the shape for forcing current local multi-view image to transform to approximate reference block image State, EtIndicate temporary stable state, EsIndicate the spatial smoothness deformation of regularization adjacent vertex, λr、λt、λsJoin for state Number, for correcting influence of three states for final energy function；
ErCalculate all feature consistencies distance and,
It is plBilinear interpolation weight；
EtIt is defined as：
Whether B indicates indicator function, for examining some pixel in static background；B(pl)=0 indicates to work as plPositioned at movement When on object,Refer to 2 dimension coordinates of the matched localview of previous frame, displacements of the s between two frames.
8. method as described in any one of claim 1 to 7, which is characterized in that the step A4 includes：
A4.1, the Color Style of any reference block is assigned to corresponding local multi-view image, transfer method is reflected according to affine color Penetrate model；
A4.2, it is directed to existing overlapping region between local multi-view image, Graphcut algorithms is applied on reference block, to replace The local multi-view image being overlapped on reference block.
CN201810183539.3A 2018-03-06 2018-03-06 A kind of 1,000,000,000 pixel video generation method of EO-1 hyperion Pending CN108429889A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810183539.3A CN108429889A (en) 2018-03-06 2018-03-06 A kind of 1,000,000,000 pixel video generation method of EO-1 hyperion

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810183539.3A CN108429889A (en) 2018-03-06 2018-03-06 A kind of 1,000,000,000 pixel video generation method of EO-1 hyperion

Publications (1)

Publication Number Publication Date
CN108429889A true CN108429889A (en) 2018-08-21

Family

ID=63157387

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810183539.3A Pending CN108429889A (en) 2018-03-06 2018-03-06 A kind of 1,000,000,000 pixel video generation method of EO-1 hyperion

Country Status (1)

Country Link
CN (1) CN108429889A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110009032A (en) * 2019-03-29 2019-07-12 江西理工大学 A kind of assembling classification method based on high light spectrum image-forming
CN111932594A (en) * 2020-09-18 2020-11-13 西安拙河安见信息科技有限公司 Billion pixel video alignment method and device based on optical flow and medium

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106339674A (en) * 2016-08-17 2017-01-18 中国地质大学（武汉） Hyperspectral image classification method based on edge preservation and graph cut model
CN106504208A (en) * 2016-10-27 2017-03-15 西京学院 Based on orderly minima and the high-spectrum image width destriping method of wavelet filtering
CN107734271A (en) * 2017-10-16 2018-02-23 深圳市未来媒体技术研究院 The pixel video generation method of HDR 1,000,000,000

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN106339674A (en) * 2016-08-17 2017-01-18 中国地质大学（武汉） Hyperspectral image classification method based on edge preservation and graph cut model
CN106504208A (en) * 2016-10-27 2017-03-15 西京学院 Based on orderly minima and the high-spectrum image width destriping method of wavelet filtering
CN107734271A (en) * 2017-10-16 2018-02-23 深圳市未来媒体技术研究院 The pixel video generation method of HDR 1,000,000,000

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110009032A (en) * 2019-03-29 2019-07-12 江西理工大学 A kind of assembling classification method based on high light spectrum image-forming
CN110009032B (en) * 2019-03-29 2022-04-26 江西理工大学 Hyperspectral imaging-based assembly classification method
CN111932594A (en) * 2020-09-18 2020-11-13 西安拙河安见信息科技有限公司 Billion pixel video alignment method and device based on optical flow and medium
CN111932594B (en) * 2020-09-18 2023-12-19 西安拙河安见信息科技有限公司 Billion pixel video alignment method and device based on optical flow and medium

Similar Documents

Publication Publication Date Title
Szeliski Video mosaics for virtual environments
JP6561216B2 (en) Generating intermediate views using optical flow
EP2383699B1 (en) Method for estimating a pose of an articulated object model
JP6201476B2 (en) Free viewpoint image capturing apparatus and method
JP4828506B2 (en) Virtual viewpoint image generation device, program, and recording medium
US11568516B2 (en) Depth-based image stitching for handling parallax
CN110211043A (en) A kind of method for registering based on grid optimization for Panorama Mosaic
CN109544447A (en) A kind of image split-joint method, device and storage medium
CN109919911A (en) Moving three dimension method for reconstructing based on multi-angle of view photometric stereo
CN106534670B (en) It is a kind of based on the panoramic video generation method for connecting firmly fish eye lens video camera group
CN113313828B (en) Three-dimensional reconstruction method and system based on single-picture intrinsic image decomposition
CN110246161B (en) Method for seamless splicing of 360-degree panoramic images
CN110211169B (en) Reconstruction method of narrow baseline parallax based on multi-scale super-pixel and phase correlation
US11922562B2 (en) Methods and systems for rendering view-dependent images using 2D images
KR20190062102A (en) Method and apparatus for operating 2d/3d augument reality technology
KR20170130150A (en) Camera rig method for acquiring 3d data, camera rig system performing the same, and storage medium storing the same
CN116958437A (en) Multi-view reconstruction method and system integrating attention mechanism
Starck et al. Virtual view synthesis of people from multiple view video sequences
CN108429889A (en) A kind of 1,000,000,000 pixel video generation method of EO-1 hyperion
Fu et al. Seamless texture optimization for RGB-D reconstruction
CN108898550A (en) Image split-joint method based on the fitting of space triangular dough sheet
Zhang et al. Pseudo-LiDAR point cloud magnification
CN109272445A (en) Panoramic video joining method based on Sphere Measurement Model
Seitner et al. Trifocal system for high-quality inter-camera mapping and virtual view synthesis
JP3540696B2 (en) Image synthesizing method, image synthesizing device, recording medium storing image synthesizing program

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication

Application publication date: 20180821

RJ01 Rejection of invention patent application after publication