WO2019062177A1 - 基于主视点的全景视频映射方法 - Google Patents

基于主视点的全景视频映射方法 Download PDF

Info

Publication number
WO2019062177A1
WO2019062177A1 PCT/CN2018/088806 CN2018088806W WO2019062177A1 WO 2019062177 A1 WO2019062177 A1 WO 2019062177A1 CN 2018088806 W CN2018088806 W CN 2018088806W WO 2019062177 A1 WO2019062177 A1 WO 2019062177A1
Authority
WO
WIPO (PCT)
Prior art keywords
angle
coordinate
spherical
center
axis
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.)
Ceased
Application number
PCT/CN2018/088806
Other languages
English (en)
French (fr)
Inventor
王荣刚
王悦名
王振宇
高文
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.)
Peking University Shenzhen Graduate School
Original Assignee
Peking University Shenzhen Graduate School
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 Peking University Shenzhen Graduate School filed Critical Peking University Shenzhen Graduate School
Priority to US16/650,141 priority Critical patent/US11301953B2/en
Publication of WO2019062177A1 publication Critical patent/WO2019062177A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T3/00Geometric image transformations in the plane of the image
    • G06T3/06Topological mapping of higher dimensional structures onto lower dimensional surfaces
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T3/00Geometric image transformations in the plane of the image
    • G06T3/12Panospheric to cylindrical image transformations
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T3/00Geometric image transformations in the plane of the image
    • G06T3/40Scaling of whole images or parts thereof, e.g. expanding or contracting
    • G06T3/4007Scaling of whole images or parts thereof, e.g. expanding or contracting based on interpolation, e.g. bilinear interpolation
    • GPHYSICS
    • G06COMPUTING OR CALCULATING; COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T3/00Geometric image transformations in the plane of the image
    • G06T3/40Scaling of whole images or parts thereof, e.g. expanding or contracting
    • G06T3/4038Image mosaicing, e.g. composing plane images from plane sub-images

Definitions

  • the present invention relates to the field of virtual reality (VR) video technology, and in particular, to a new primary view-based panoramic video mapping method and a corresponding inverse mapping method, which is used for panoramic video to ensure the quality of the main view area of the panoramic video.
  • VR virtual reality
  • the file size of the panoramic video and the code rate of the encoding can be greatly reduced.
  • the virtual reality video provides a 360-degree view
  • the user only views an image near a viewpoint for a short period of time.
  • This viewpoint is called the main viewpoint viewed by the user.
  • asymmetric mapping technology and stream switching technology It was proposed to achieve the purpose of saving code rate.
  • the asymmetric mapping technique uses a higher sampling density to sample the main view area and a lower area to sample with a lower sampling density.
  • the stream switching technology encodes and stores the code stream of the asymmetric mapping video with different main points of view on the server side, and the client selects and receives the corresponding stream according to the location of the user's viewpoint.
  • the asymmetric mapping technology and the stream switching technology need to improve the storage and encoding cost of the server, the transmission cost of the network and the decoding complexity of the client can be effectively reduced, which is very effective in a one-to-many application scenario.
  • the coding efficiency of the existing asymmetric mapping technology is still low, and needs to be improved.
  • the present invention proposes a new panoramic video asymmetric mapping method and a corresponding inverse mapping method.
  • the panoramic video asymmetric mapping technology greatly reduces the resolution of the rest of the video while ensuring that the video quality of the main view area is constant, thereby effectively saving the code rate required for virtual reality video coding.
  • the panoramic video asymmetric anti-mapping technology provides a method of mapping back to the spherical surface from a plane, by which the planar video can be mapped back to the spherical surface for rendering and viewing.
  • the farther area uses a lower sampling density to save the code rate; further, the passing center O establishes a right-handed coordinate system, the Z-axis points to the direction of the spherical longitude and latitude (0°, 0°), and the Y-axis points to the spherical north pole.
  • the direction of the X axis points to the direction of the spherical longitude and latitude (90°, 0°); a quadrangular pyramid W is created, the bottom of which is centered on the positive half of the Z axis, and the apex is on the negative half of the Z axis.
  • the line connecting the center D of the bottom surface of the quadrangular pyramid to the center O is denoted by l 1
  • the line connecting the center of the bottom side of the quadrangular pyramid to the center O is denoted as l 2
  • l 1 and l angle is an angle of 2 ⁇
  • the bottom surface of the quadrangular pyramid I the side intersecting the positive half-axis of the X-axis is referred to as face II
  • the side intersecting the positive half-axis of the Y-axis is referred to as face III
  • the side intersecting the negative half-axis of the X-axis is referred to as face IV
  • Y The side where the negative half-axis intersects is denoted as face V"; when the pitch of the main viewpoint, the yaw and the roll are ( ⁇ 1 , ⁇ 2 , ⁇ 3 ) respectively, the corresponding four
  • the pyramid is a quadrangular
  • the mapping method of the present invention maps the above region I to a rectangular plane having a resolution of W I' ⁇ H I '
  • the regions II, III, IV, and V are mapped to the four isosceles triangle planes II', III', IV', and V', respectively, and the bottom and height of the four isosceles triangle planes are respectively Four isosceles triangles are then spliced into a rectangular plane VI' with a resolution of W II ' ⁇ H II '.
  • the parameters ⁇ , W I′ , H I′ , W II′ , H II′ , ⁇ 1 , ⁇ 2 , ⁇ 3 are all set by themselves; the rectangular plane I′ and the rectangular plane VI′ are the two-dimensional images obtained by mapping. Or video B.
  • the mapping format of the above panoramic image A includes, but is not limited to, a warp and a latitude map, a cube-mapped image, and a panoramic video captured by a multi-way camera.
  • the above two-dimensional image or video B includes the above-described rectangular plane I' and rectangular plane VI'.
  • the method of mapping the spherical surface corresponding to the panoramic image or the video A to the two-dimensional planar image or the video B includes the following steps:
  • the first step for each pixel in the rectangular plane I', according to its coordinates (X I' , Y I' ) in the plane I', calculate the coordinate Coordinate' corresponding to the bottom surface of the quadrangular pyramid Q', Then, according to the method of perspective projection, the coordinate Coordinate corresponding to the spherical surface is calculated. Finally, the pixel value of the corresponding position on the spherical surface is taken according to the spherical coordinate Coordinate (or the corresponding pixel value is calculated by interpolation in the vicinity of the pixel), as the pixel point in the plane I′. (X I, Y I) of the pixel values, the step of calculating the specific spherical coordinate plane coordinates according coordinate (X I ', Y I' ) is:
  • Step 2 For each pixel in the four isosceles triangle planes II', III', IV', V', according to its coordinates (X II' , Y II' ) (X and Y axes respectively vertical and Parallel to the bottom edge of the isosceles triangle), calculate the coordinate Coordinate' corresponding to the side of the quadrangular pyramid Q', and then further calculate the coordinate Coordinate corresponding to the spherical surface according to the perspective projection method, and finally take the spherical surface according to the spherical coordinate Coordinate
  • the pixel value of the corresponding position (or the nearby pixel is calculated by interpolation), as the pixel value of the pixel point (X II ' , Y II ' ), the spherical coordinate Coordinate is calculated according to the plane coordinate (X II ' , Y II ' )
  • the specific steps are:
  • the point at which the pixel point (X II' , Y II' ) corresponds to the side of the quadrangular pyramid Q′ is denoted as A II′ , and the line connecting the center and the apex of the side edge of the quadrangular pyramid Q′ is denoted as n 1 , four
  • the line connecting the apex of the pyramid Q' and the center of the bottom surface is denoted by n 2
  • the projections of the point A II' on n 1 and n 2 are denoted as B II ' and C II ' respectively
  • the center of the bottom surface of the quadrangular pyramid Q' is denoted as D
  • a The orientation of II' can be determined by the angle B II' OD and the angle A II' C II' B II' ; in the present invention, the angle B II' OD and the angle A are calculated from the values of (X II' , Y II' ).
  • the size of II' C II' B II' is a function of the value of Y II' .
  • the relationship between the size of the angle B II' OD and X II ' is ⁇ B.
  • the third step splicing the four isosceles triangular planes II', III', IV', V' obtained in the second step into a rectangular plane VI' with a resolution of W II' ⁇ H II '.
  • the rectangular plane I' and the rectangular plane VI' are the two-dimensional images or videos B obtained by the mapping.
  • f(X II' ) may be:
  • panoramic video demapping process is based on the primary view, the two-dimensional image or video back spherical mapping process B;
  • B comprises a two-dimensional image or video resolution W I ' ⁇ H I' rectangular plane I 'and
  • the resolution is W II' ⁇ H II' rectangular plane VI', and the rectangular plane VI' can be further divided into four isosceles triangle planes II', III', IV', V', and the bottom and height of four isosceles triangles
  • the demapping method first maps the plane to an equilateral pyramid by means of isometric projection, and then maps the quadrilateral pyramid to the spherical surface.
  • the passing center O establishes a right-handed coordinate system, the Z-axis points to the direction of the spherical longitude and latitude (0°, 0°), the Y-axis points to the direction of the spherical north pole, and the X-axis points to the spherical longitude and latitude (90°, 0°); a quadrangular pyramid W is created, the bottom of which is centered on the positive half of the Z axis, the apex is on the negative half of the Z axis, and the bottom side is parallel to the X and Y axes, respectively.
  • the line connecting the center O is denoted by l 1
  • the line connecting the center of the bottom side of the quadrangular pyramid to the center O is denoted by l 2
  • the angle formed by l 1 and l 2 is ⁇
  • represents the main view area
  • the size is such that the bottom surface of the quadrangular pyramid is referred to as the surface I", the side intersecting the positive half-axis of the X-axis is referred to as the surface II", and the side intersecting the positive half-axis of the Y-axis is referred to as the surface III", and the negative half-axis of the X-axis
  • the intersecting side is denoted as face IV
  • the side intersecting the negative half of the Y-axis is denoted as face V";
  • the four sides of the spherical center O and the quadrangular pyramid Q' are divided into four fan-shaped planes, and the four planes further divide the non-primary view region into four sub-regions, and the four sub-regions respectively correspond to the four sides of the pyramid II", III", IV ", V", respectively, is referred to as regions II, III, IV, V.
  • the inverse mapping method of the present invention maps a rectangular plane I' of a resolution of W I' ⁇ H I' contained in a two-dimensional image or video B to a spherical surface
  • the four isosceles triangle planes II', III', IV', V' included in the two-dimensional image or video B are mapped onto the non-main viewpoint areas II, III, IV, V of the spherical surface.
  • the values of the parameters ⁇ , W I' , H I' , W II ' , H II ' , ⁇ 1 , ⁇ 2 , ⁇ 3 include, but are not limited to, those obtained from the code stream.
  • the method of mapping the planar image or the video B back to the spherical surface is to perform the following operations on all points on the spherical surface:
  • the first step judging which one of the regions I, II, III, IV, V is located according to the coordinate Coordinate of the spherical point and the values of ⁇ 1 , ⁇ 2 , ⁇ 3 ; if it is located in the region I, jump to The second step; if it is located in the area II, III, IV, V, then skip to the fifth step;
  • the second step the point where the spherical coordinate is Coordinate is denoted as A I′ , the line connecting the center of the quadrangular pyramid Q′ to the edge is recorded as m 1 and m 2 , and the projection of the point A I′ on m 1 and m 2 It is denoted as B I' and C I' respectively , and the center of the bottom surface of the quadrangular pyramid Q' is denoted as D; the angle B I' OD and the angle C I are calculated according to the coordinates Coordinate of A I' and the values of ⁇ 1 , ⁇ 2 , ⁇ 3 ' The size of the OD;
  • the third step calculating the values of the plane coordinates (X I' , Y I' ) according to the magnitudes of the angles B I′ OD and the angle C I′ OD, the values of X I′ and Y I′ and the angle B I′ OD , respectively And the size of the angle C I' OD is a functional relationship;
  • the fourth step taking the pixel value (or nearby pixels interpolated) on the rectangular plane I' (X I' , Y I' ) as the pixel value of the point on the spherical coordinate Coordinate; skip the subsequent steps;
  • Step 5 Record the point where the spherical coordinate is Coordinate as A II' , the line connecting the center and the apex of the side edge of the quadrilateral pyramid Q' as n 1 , and the plane of n 1 to the center of the spherical center O as ⁇ 3 , four
  • the line connecting the apex of the pyramid Q' and the center of the bottom surface is denoted by n 2
  • the projection of the point A II' on ⁇ 3 and n 2 is denoted as B II ' and C II ' respectively
  • the center of the bottom surface of the quadrangular pyramid Q' is denoted as D
  • the coordinates of A II ' and the values of ⁇ 1 , ⁇ 2 , ⁇ 3 are used to calculate the size of the angle B II′ OD and the angle A II′ C II′ B II′ ;
  • Step 6 Calculate the values of the plane coordinates (X II' , Y II' ) according to the size of the angle B II′ OD and the angle A II′ C II′ B II′ (the X and Y coordinate axes are perpendicular and parallel to each other, etc.)
  • the value of Y II' is a function of the magnitude of the angle A II' C II' B II' .
  • Step 7 taking the pixel value (or nearby pixels) on the triangular plane (X II ' , Y II ' ) as the pixel value of the point on the spherical surface Co coordinate;
  • the first to seventh operations are performed on all the points on the spherical surface, thereby obtaining a panoramic image of the spherical surface.
  • f(X II' ) may be:
  • the invention provides a new panoramic video asymmetric mapping method and a corresponding inverse mapping method, and the spherical surface corresponding to the panoramic image or the video A is mapped to the two-dimensional image or the video B by the mapping method; the spherical surface is first projected onto the bottom surface On the square isosceles quadrilateral pyramid, the quadrangular pyramid is further projected onto the plane; the projection of the area of the main viewpoint uses the isometric projection and uses a higher sampling density to ensure that the video quality of the main viewpoint area is higher, The non-primary view area uses a lower sampling density to save the code rate.
  • the panoramic video asymmetric mapping technology greatly reduces the resolution of the rest of the video while ensuring that the video quality of the main view area is constant, thereby effectively saving the code rate required for virtual reality video coding.
  • the panoramic video asymmetric anti-mapping technology provides a method of mapping back to the spherical surface from a plane, by which the planar video can be mapped back to the spherical surface for rendering and viewing.
  • the invention overcomes the deficiencies of the prior art and further improves the coding efficiency of the asymmetric mapping technique, and has the following advantages:
  • the relationship between the angle on the spherical surface of the main viewpoint area and the plane coordinate is a linear function, that is, the sampling of the main viewpoint according to the angle of the equal angle ensures that the sampling of the main viewpoint area is relatively uniform;
  • the panoramic image mapping parameter ⁇ and the function f in the present invention are adjustable, that is, the range of the main viewpoint area and the variation speed of the sampling density are adjustable;
  • planar image B in the present invention can be mapped back to the spherical surface for rendering viewing.
  • FIG. 1 is a schematic diagram showing a relationship between a spherical surface and a plane in the mapping process of the present invention
  • (a) is a schematic diagram of the corresponding quadrangular pyramid and spherical surface when the pitch angle of the main viewpoint, the yaw angle and the roll angle are (0°, 0°, 0°);
  • b) is a schematic diagram of the corresponding quadrangular pyramid and spherical surface when the pitch angle of the main viewpoint, the yaw angle and the roll angle are ( ⁇ 1 , ⁇ 2 , ⁇ 3 ), respectively;
  • (c) A schematic diagram of a rectangular plane I' and a rectangular plane VI' obtained by mapping a spherical surface to a plane by the method of the present invention;
  • O is the center of the sphere; the Z axis points to the direction of the spherical longitude and latitude (0°, 0°), the Y axis points to the direction of the spherical north pole, and the X axis points to the spherical longitude and latitude (90°, 0°)
  • the angle between the center of the line and the bottom side of the quadrangular pyramid and the line connecting the center O is ⁇ , and ⁇ represents the size of the main view area;
  • a quadrangular pyramid is a result of rotating the quadrangular pyramid in (a) according to a pitch, a yaw, and a roll around the center O;
  • I' is a rectangular plane; II', III', IV', V' are four isosceles triangle planes, II', III', IV', V' are spliced into a rectangular plane VI'
  • FIG. 2 is a schematic diagram showing a mapping relationship between a point of a coordinate plane ( I I' , Y I' ) and a point A I' on a quadrangular pyramid in the rectangular plane I' of the present invention
  • (a) is a schematic diagram of a point where the coordinates of the rectangular face I' are (X I' , Y I' );
  • (b) is a schematic view of the point A I' on the quadrangular pyramid; in the figure, O is the center of the sphere D is the center of the main viewpoint, m 1 and m 2 are the lines connecting the midpoint of the bottom of the quadrangular pyramid, respectively, and the point A I' is the point of the coordinate of the rectangular plane I' (X I' , Y I' ) Mapped to corresponding points on the pyramid, B I' and C I' are projections of points A I' on m 1 and m 2 , respectively.
  • Figure 3 is a schematic diagram showing the mapping relationship between a point of coordinates (X II' , Y II ' ) and a point A II' on a quadrangular pyramid in a triangular plane in the present invention
  • (a) is a schematic diagram of a point in the triangular plane with coordinates (X II' , Y II' );
  • (b) is a schematic diagram of a point A II' on a quadrangular pyramid; in the figure, O is the center of the sphere, and D is At the center of the main viewpoint, n 1 is the line connecting the center and the apex of the bottom side of the quadrangular pyramid, n 2 is the line connecting the apex of the quadrangular pyramid and the center D of the bottom surface, and the point A II′ is the coordinate in the plane of the triangle (X II′ , Y II The points of ' ) are mapped to corresponding points on the quadrangular pyramid, and B II' and C II' are projections of points A II' on n 1 and n 2 , respectively.
  • FIG. 4 is an effect diagram of an embodiment of a panoramic image mapping method of the present invention.
  • FIG. 5 is a schematic diagram of calculating a corner BOD and an angle COD in the panoramic image demapping method of the present invention
  • point B and point C are point A on the ZOX plane and the ZOY plane, respectively; D is a point on the positive half of the Z axis, and E is a point on the negative half axis of the Z axis.
  • FIG. 6 is a schematic diagram showing a mapping relationship between a point A I' located in a region I on a spherical surface and a point having a coordinate of (X I' , Y I' ) in a rectangular plane I′ in the panoramic image demapping method of the present invention
  • (a) is a schematic view of the point A I' in the region I on the spherical surface, O is the center of the sphere, D is the center of the main viewpoint, and m 1 and m 2 are the lines connecting the midpoint of the bottom surface of the quadrangular pyramid, respectively.
  • B I' and C I' are projections of points A I' on m 1 and m 2 , respectively;
  • (b) are schematic views of points in the rectangular plane I' at coordinates (X I' , Y I' ).
  • FIG. 7 is a schematic diagram showing a mapping relationship between points A II′ located in regions II, III, IV, and V and points (X II′ , Y II′ ) in a triangular plane on a spherical surface in the panoramic image demapping method of the present invention
  • (a) is a schematic view of the point A II' in the region II, III, IV, V on the spherical surface
  • O is the center of the sphere
  • D is the center of the main viewpoint
  • n 1 is the center of the base of the quadrangular pyramid and the apex of the vertex Line
  • n 2 is the line connecting the apex of the quadrangular pyramid and the center D of the bottom surface
  • B II' and C II' are projections of points A II' on n 1 and n 2 respectively
  • (b) are points in the plane of the triangle
  • (X) Schematic representation of II' , Y II' ).
  • Embodiments of the present invention provide a method for mapping a panoramic image based on a primary view, including a panoramic image mapping method and a corresponding inverse mapping method.
  • Embodiments of the mapping method and the inverse mapping method are respectively introduced below.
  • the panoramic image mapping method we map the spherical surface corresponding to the panoramic image A (such as the latitude and longitude image, the cube mapping image, etc.) of a certain mapping format to the planar image B corresponding to the panoramic mapping designed in the present invention, and the plane Image B contains a rectangular plane I' having a resolution W I' ⁇ H I ' and a rectangular plane VI ' having a resolution W II ' ⁇ H II '.
  • the panoramic image mapping method firstly projects the spherical surface onto the isosceles pyramid having a square bottom surface, and further projects the quadrangular pyramid onto the plane; the projection uses the isometric projection on the region of the main viewpoint and uses a higher sampling density to ensure the main The video quality of the area of the viewpoint is higher, and the lower sampling density is used for the non-main viewpoint area to save the code rate. As shown in Fig.
  • the spherical center O establishes a right-handed coordinate system, the Z-axis points to the direction of the spherical longitude and latitude (0°, 0°), the Y-axis points to the direction of the spherical north pole, and the X-axis points to the spherical longitude and latitude ( 90°, 0°) direction; then create a quadrangular pyramid W whose center is on the positive half of the Z axis, the apex is on the negative half of the Z axis, and the bottom side is parallel to the X and Y axes, respectively.
  • the line connecting the center D of the bottom surface and the center of the sphere O is denoted as l 1
  • the line connecting the center of the bottom side of the quadrangular pyramid to the center of the sphere O is denoted by l 2
  • the angle formed by the angles of l 1 and l 2 is ⁇
  • represents
  • the size of the main viewpoint area is defined as the surface I′′ of the quadrangular pyramid
  • the side intersecting the positive semi-axis of the X-axis is referred to as the surface II′′
  • the side intersecting the positive semi-axis of the Y-axis is referred to as the surface III′′
  • the side where the negative half-axis intersects is denoted as the face IV", and the side intersected with the negative half-axis of the Y-axis is denoted as the face V";
  • the pitch of the main viewpoint the yaw and the roll are respectively ( ⁇ 1, ⁇ 2, ⁇ 3) corresponding to the quadrang
  • the four sides of the spherical center O and the quadrangular pyramid Q' are divided into four fan-shaped planes, and the four planes further divide the non-main viewpoint area into four sub-areas, and the four sub-areas respectively correspond to the four sides II", III" of the pyramid.
  • IV", V respectively referred to as regions II, III, IV, V.
  • the mapping method of the present invention maps the above region I to a rectangular plane I' having a resolution of W I' ⁇ H I ' , and the region II , III, IV, V are respectively mapped to the four isosceles triangle planes II', III', IV', V', and the bottom and height of the four isosceles triangle planes are respectively Four isosceles triangles are then spliced into a rectangular plane VI' with a resolution of W II ' ⁇ H II '.
  • the parameters ⁇ , W I' , H I' , W II ' , H II ' , ⁇ 1 , ⁇ 2 , ⁇ 3 are all set by themselves.
  • the specific process of the panoramic image mapping method is as follows:
  • the first step for each pixel in the rectangular plane I', according to its coordinates (X I' , Y I' ) in the plane I', calculate the coordinate Coordinate' corresponding to the bottom surface of the quadrangular pyramid Q', Then, according to the method of perspective projection, the coordinate Coordinate corresponding to the spherical surface is calculated. Finally, the pixel value of the corresponding position on the spherical surface is taken according to the spherical coordinate Coordinate (or the corresponding pixel value is calculated by interpolation in the vicinity of the pixel), as the pixel point in the plane I′. (X I, Y I) of the pixel values, the step of calculating the specific spherical coordinate plane coordinates according coordinate (X I ', Y I' ) is:
  • the point of the plane I' with the coordinates (X I' , Y I' ) corresponds to the point on the bottom surface of the quadrangular pyramid Q′ is A I′
  • the bottom of the quadrangular pyramid Q′ is at the center of the edge
  • the lines are denoted as m 1 and m 2
  • the projections of points A I' on m 1 and m 2 are B I' and C I' , respectively
  • the angle B I is calculated from the value of (X I' , Y I' ).
  • the size of ' OD and angle C I' OD which is calculated as:
  • T is a 3 ⁇ 3 dimensional rotation matrix generated from the pitch angle ⁇ 1 , the yaw angle ⁇ 2 and the roll angle ⁇ 3 ;
  • k is the length of the line segment OD;
  • R is the radius of the ball; It represents OA I 'of the mold segment, i.e. OA I' length of the segment, Divide by That is right Perform normalization.
  • step (1.2) the value of k in the above step (1.2) is unknown, but does not affect the subsequent calculation, because in step (1.3) When normalized, k will be eliminated.
  • Step 2 For each pixel in the four isosceles triangle planes II', III', IV', V', according to its coordinates (X II' , Y II' ) (X and Y axes respectively vertical and Parallel to the bottom edge of the isosceles triangle), calculate the coordinate Coordinate' corresponding to the side of the quadrangular pyramid Q', and then further calculate the coordinate Coordinate corresponding to the spherical surface according to the perspective projection method, and finally take the spherical surface according to the spherical coordinate Coordinate
  • the pixel value of the corresponding position (or the nearby pixel is calculated by interpolation), as the pixel value of the pixel point (X II ' , Y II ' ), the spherical coordinate Coordinate is calculated according to the plane coordinate (X II ' , Y II ' )
  • the specific steps are:
  • f function is any function that satisfies the following conditions:
  • H is the height of the isosceles triangle
  • L is the base of the isosceles triangle
  • W II ' W II ' , H II ' , W II ' , H II ' ;
  • T is a 3 ⁇ 3 dimensional rotation matrix generated from the pitch angle ⁇ 1 , the yaw angle ⁇ 2 and the roll angle ⁇ 3 ;
  • k is the length of the line segment C II′ B II′ ;
  • R is the radius of the ball; It represents OA I 'of the mold segment, i.e. OA I' length of the segment, Divide by That is right Perform normalization.
  • step (2.2) the value of k in the above step (2.2) is unknown, but does not affect the subsequent calculation, because in step (2.3) When normalized, k will be eliminated.
  • the third step splicing the four isosceles triangle planes II', III', IV', V' obtained in the second step into a rectangular plane VI' with a resolution of W II' ⁇ H II ';
  • a two-dimensional image or video B is mapped back to the spherical surface based on the main viewpoint;
  • the two-dimensional image or video B contains a rectangular plane having a resolution of W I' ⁇ H I ' I' and the resolution are W II' ⁇ H II' rectangular plane VI', and the rectangular plane VI' can be further divided into four isosceles triangle planes II', III', IV', V', four isosceles triangles Bottom and height are
  • the demapping method first maps the plane to an equilateral pyramid by means of isometric projection, and then maps the quadrilateral pyramid to the spherical surface.
  • the passing center O establishes a right-handed coordinate system, the Z-axis points to the direction of the spherical longitude and latitude (0°, 0°), the Y-axis points to the direction of the spherical north pole, and the X-axis points to the spherical longitude and latitude (90°, 0°); a quadrangular pyramid W is created, the bottom of which is centered on the positive half of the Z axis, the apex is on the negative half of the Z axis, and the bottom side is parallel to the X and Y axes, respectively.
  • the line connecting the center O is denoted by l 1
  • the line connecting the center of the bottom side of the quadrangular pyramid to the center O is denoted by l 2
  • the angle formed by l 1 and l 2 is ⁇
  • represents the main view area
  • the size is such that the bottom surface of the quadrangular pyramid is referred to as the surface I", the side intersecting the positive half-axis of the X-axis is referred to as the surface II", and the side intersecting the positive half-axis of the Y-axis is referred to as the surface III", and the negative half-axis of the X-axis
  • the intersecting side is denoted as face IV
  • the side intersecting the negative half of the Y-axis is denoted as face V";
  • the four sides of the spherical center O and the quadrangular pyramid Q' are divided into four fan-shaped planes, and the four planes further divide the non-primary view region into four sub-regions, and the four sub-regions respectively correspond to the four sides of the pyramid II", III", IV ", V", respectively referred to as regions II, III, IV, V.
  • the demapping method of the present invention maps a rectangular plane I' of a resolution of W I' ⁇ H I ' contained in a two-dimensional image or video B to a spherical surface
  • the four isosceles triangle planes II', III', IV', V' included in the two-dimensional image or video B are mapped onto the non-main viewpoint areas II, III, IV, V of the spherical surface.
  • the values of the parameters ⁇ , W I' , H I' , W II ' , H II ' , ⁇ 1 , ⁇ 2 , ⁇ 3 include, but are not limited to, are obtained from a code stream.
  • the specific process of the panoramic image back mapping method is as follows :
  • the first step judging which one of the regions I, II, III, IV, V is located according to the coordinate Coordinate of the spherical point and the values of ⁇ 1 , ⁇ 2 , ⁇ 3 ; the specific method is:
  • the coordinate Coordinate of the spherical point is rotated.
  • the coordinate Coordinate is represented by Cartesian coordinates as (X ball , Y ball , Z ball ), and the rotated point is recorded as A, A coordinates. for:
  • T' is a 3 ⁇ 3 dimensional rotation matrix generated from the pitch angle - ⁇ 1 , the yaw angle - ⁇ 2 and the roll angle - ⁇ 3 ;
  • point A is projected onto the ZOX plane and the ZOY plane, and the projection points are point B and point C, respectively, and the angles of the angle BOD and the angle COD are calculated:
  • step 2 If it is in area I, skip to step 2; if it is in area II, III, IV, V, skip to step 5;
  • the second step as shown in Fig. 6, the point where the spherical coordinate is Coordinate is denoted as A I' , the coordinate Coordinate is represented by Cartesian coordinates as (X ball , Y ball , Z ball ), and the bottom face of the quadrangular pyramid Q'
  • the lines at the center of the side are denoted by m 1 and m 2
  • the projections of points A I' on m 1 and m 2 are denoted as B I' and C I' , respectively
  • the center of the bottom of the quadrangular pyramid Q' is denoted as D; according to A I 'coordinate coordinate and ⁇ 1, ⁇ 2, ⁇ 3 to calculate the angle value B I' OD size and angle C I 'OD of:
  • T' is a 3 ⁇ 3 dimensional rotation matrix generated from the pitch angle - ⁇ 1 , the yaw angle - ⁇ 2 and the roll angle - ⁇ 3 ;
  • the calculation of the angle B I' OD and the angle C I' OD is the same as the calculation method of the angle BOD and the angle COD in the first step. Therefore, the value of the first-order mid-angle BOD and the angle COD can also be directly used as the angle B I'. OD and the value of the angle C I' OD;
  • the third step calculating the values of the plane coordinates (X I' , Y I' ) according to the angle B I' OD and the angle C I' OD, the calculation formula is as follows:
  • the fourth step taking the pixel value (or nearby pixels interpolated) on the rectangular plane I' (X I' , Y I' ) as the pixel value of the point on the spherical coordinate Coordinate; skip the subsequent steps;
  • Step 5 As shown in Fig. 7, the point where the spherical coordinate is Coordinate is denoted as A II' , the coordinate Coordinate is represented by Cartesian coordinates as (X ball , Y ball , Z ball ), and the quadrangular pyramid Q' side bottom edge
  • the line connecting the center and the apex is denoted by n 1
  • the plane passing n 1 to the center of the sphere O is denoted by ⁇ 2
  • the line connecting the apex of the quadrangular pyramid Q′ and the center of the bottom surface is denoted by n 2
  • the point A II′ is at ⁇ 3 and n
  • the projections on 2 are denoted as B II' and C II' , respectively, and the center of the bottom of the quadrangular pyramid Q' is denoted as D
  • the angle B II' OD is calculated according to the coordinates Coordinate of A II' and the values of ⁇ 1 , ⁇ 2 , ⁇ 3
  • T' is a 3 ⁇ 3 dimensional rotation matrix generated from the pitch angle - ⁇ 1 , the yaw angle - ⁇ 2 and the roll angle - ⁇ 3 ;
  • Step 6 Calculate the values of the plane coordinates (X II' , Y II' ) according to the size of the angle B II′ OD and the angle A II′ C II′ B II′ (the X and Y coordinate axes are perpendicular and parallel to each other, etc.)
  • the relationship between the value of Y II' and the angle A II' C II' B II' is:
  • Step 7 taking the pixel value (or interpolation of nearby pixels) on the corresponding triangle plane (X II ' , Y II ' ) as the pixel value of the point on the spherical surface Co coordinate;

Landscapes

  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Image Processing (AREA)

Abstract

一种全景视频非对称映射方法及相应的反映射方法,通过映射方法将全景图像或视频A所对应的球面映射到二维图像或视频B上;先将球面投影到底面为方形的等腰四棱锥上,再进一步将四棱锥投影到平面上;投影中对主视点的区域使用等角投影并使用较高的采样密度,保证主视点的区域的视频质量较高,对非主视点区域使用较低的采样密度以节省码率。该全景视频非对称映射技术在保证主视点区域视频质量不变的情况下,大大降低视频其余区域的分辨率,有效地节省了虚拟现实视频编码所需的码率。全景视频非对称反映射技术提供了从平面映射回球面的方法,通过此方法可将平面视频映射回球面进行渲染观看。

Description

基于主视点的全景视频映射方法 技术领域
本发明涉及虚拟现实(VR)视频技术领域,尤其涉及一种新的基于主视点的全景视频映射方法及相应的反映射方法,该映射技术用于全景视频,在保证全景视频的主视点区域质量不变的前提下,可大大减少全景视频的文件大小和编码的码率。
背景技术
随着虚拟现实视频的发展,对虚拟现实视频的需求日益增加。相较于传统视屏100多度的视角,虚拟现实视频需要提供360度的视角,因此虚拟现实视频需要更大的分辨率,此外虚拟现实视频通常需要在头戴式设备上观看,为了不给用户造成眩晕感,建议的帧率需要在60帧每秒以上,甚至需要120帧每秒。由于分辨率和帧率的增加,虚拟现实视频编码所需的码率也比传统视频大大增加。
虽然虚拟现实视频提供了360度的视角,但是用户在一小段时间内只会观看一个视点附近的图像,这个视点方向称为用户观看的主视点,针对这个特点,非对称映射技术和流切换技术被提出以实现节省码率的目的。非对称映射技术是使用较高的采样密度对主视点区域进行采样,而对别的区域使用较低的采样密度进行采样。流切换技术在服务器端编码存储多路主视点不同的非对称映射视频的码流,客户端根据用户视点的位置选择接收对应的流。
使用非对称映射技术和流切换技术虽然需要提高服务器端的存储和编码成本,但是能有效地降低网络的传输成本和客户端的解码复杂度,在一对多的应用场景下十分有效。但现有的非对称映射技术的编码效率较低下,还有待提高。
发明内容
为了克服上述现有技术的不足,进一步提高非对称映射技术的编码效率,本发明提出了一种新的全景视频非对称映射方法及相应的反映射方法。该全景视频非对称映射技术在保证主视点区域视频质量不变的情况下,大大降低视频其余区域的分辨率,有效地节省了虚拟现实视频编码所需的码率。全景视频非对称反映射技术提供了从平面映射回球面的方法,通过此方法可将平面视频映射回球面进行渲染观看。
本发明提供的技术方案是:
一种新的基于主视点的全景视频非对称映射方法,将全景图像或视频A所对应的球面映 射到二维图像或视频B上;所述映射方法先将球面投影到底面为方形的等腰四棱锥上,再进一步将四棱锥投影到平面上;投影中对主视点附近的区域使用等角投影并使用较高的采样密度,保证主视点附近的区域的视频质量较高,对距离主视点较远的区域使用较低的采样密度以节省码率;进一步的,过球心O建立右手坐标系,Z轴指向球面经度和纬度为(0°,0°)的方向,Y轴指向球面北极的方向,X轴指向球面经度和纬度为(90°,0°)的方向;再建立一个四棱锥W,其底面中心在Z轴正半轴上,顶点在Z轴负半轴上,底面边分别与X轴和Y轴平行,四棱锥的底面中心D与球心O的连线记作l 1,四棱锥的底面边的中心与球心O的连线记作l 2,l 1和l 2所成角的角度为θ,将四棱锥的底面记为面I″,与X轴正半轴相交的侧面记为面II″,与Y轴正半轴相交的侧面记为面III″,与X轴负半轴相交的侧面记为面IV″,与Y轴负半轴相交的侧面记为面V″;当主视点的俯仰角(pitch),航偏角(yaw)和翻滚角(roll)分别为(β 1,β 2,β 3)时对应的四棱锥为上述四棱锥Q绕球心O旋转(β 1,β 2,β 3)后对应的四棱锥Q′,过球心O与四棱锥Q′的四条底边作四个扇形平面,四个扇形平面将球面划为两个区域,其中包含主视点的区域称为主视点区域,其对应棱锥的底面,记为区域I,另一区域称为非主视点区域,过球心O与四棱锥Q′的四条侧边作四个扇形平面,四个平面进一步将非主视点区域划分为四个子区域,四个子区域分别对应棱锥的四个侧面II″,III″,IV″,V″,分别记作区域II、III、IV、V。本发明的映射方法将上述的区域I映射到分辨率为W I′×H I′的矩形平面I′上,将区域II、III、IV、V分别映射到四个等腰三角形平面II′,III′,IV′,V′上,四个等腰三角形平面的底和高分别为
Figure PCTCN2018088806-appb-000001
再将四个等腰三角形拼接成分辨率为W II′×H II′矩形平面VI′。所述参数θ,W I′,H I′,W II′,H II′,β 1,β 2,β 3均自行设置;矩形平面I′和矩形平面VI′即为映射得到的二维图像或视频B。
上述全景图像A的映射格式包括但不限于经纬图、立方体映射图像、多路相机采集的全景视频。上述二维图像或视频B包含上述矩形平面I′和矩形平面VI′。
针对上述全景图像映射方法,进一步地,将全景图像或视频A所对应的球面映射到二维平面图像或视频B上的方法包括如下步骤:
第一步:对矩形平面I′中的每个像素点,根据其在平面I′中的坐标(X I′,Y I′),计算其对应到四棱锥Q′底面上的坐标Coordinate′,然后进一步根据透视投影的方法计算其对应到球面上 的坐标Coordinate,最后根据球面坐标Coordinate取球面上对应位置的像素值(或附近像素通过插值计算得到相应像素值),作为平面I′中像素点(X I,Y I)的像素值,根据平面坐标(X I′,Y I′)计算球面坐标Coordinate的具体步骤为:
(1.1)将平面I′中坐标为(X I′,Y I′)的点对应四棱锥Q′底面上的点记为A I′,四棱锥Q′底面对边中心的连线记为m 1和m 2,点A I′在m 1和m 2上的投影分别记作B I′和C I′,四棱锥Q′底面中心记为D,A I′的方位可以由角B I′OD和角C I′OD来确定;本发明中根据(X I′,Y I′)的值来计算角B I′OD和角C I′OD的大小,角B I′OD和角C I′OD的大小分别与X I′和Y I′的值成一次函数关系;
(1.2)根据角B I′OD和角C I′OD的大小以及β 1,β 2,β 3的值求出A I′的坐标Coordinate′;
(1.3)根据点A I′的坐标,求出射线
Figure PCTCN2018088806-appb-000002
与球面的交点坐标Coordinate;
第二步:对四个等腰三角形平面II′,III′,IV′,V′中的每个像素点,根据其坐标(X II′,Y II′)(X和Y坐标轴分别垂直和平行于等腰三角形的底边),计算其对应到四棱锥Q′侧面上的坐标Coordinate′,然后进一步根据透视投影的方法计算其对应到球面上的坐标Coordinate,最后根据球面坐标Coordinate取球面上对应位置的像素值(或附近像素通过插值计算得到相应像素值),作为像素点(X II′,Y II′)的像素值,根据平面坐标(X II′,Y II′)计算球面坐标Coordinate的具体步骤为:
(2.1)将像素点(X II′,Y II′)对应到四棱锥Q′侧面上的点记为A II′,四棱锥Q′侧面底边中心和顶点的连线记为n 1,四棱锥Q′顶点和底面中心的连线记为n 2,点A II′在n 1和n 2上的投影分别记作B II′和C II′,四棱锥Q′底面中心记为D,A II′的方位可以由角B II′OD和角A II′C II′B II′来确定;本发明中根据(X II′,Y II′)的值来计算角B II′OD和角A II′C II′B II′的大小,角A II′C II′B II′的大小与Y II′的值成一次函数关系,角B II′OD的大小与X II′的关系为∠B II′OD=f(X II′),f函数为满足以下条件的任意函数:
θ=f(0)
180°=f(H)
其中,H为等腰三角形的高,对于四个等腰三角形平面II′,III′,IV′,V′,其值分别为
Figure PCTCN2018088806-appb-000003
(2.2)根据角B II′OD和角A II′C II′B II′的大小以及β 1,β 2,β 3的值求出A II′的坐标Coordinate′;
(2.3)根据点A II′的坐标,求出射线
Figure PCTCN2018088806-appb-000004
与球面的交点坐标Coordinate;
第三步:将第二步中得到的四个等腰三角形平面II′,III′,IV′,V′拼接成分辨率为W II′×H II′矩形平面VI′。
矩形平面I′和矩形平面VI′即为映射得到的二维图像或视频B。
针对上述全景视频映射过程,进一步的,步骤(2.1)中,f(X II′)可为:
Figure PCTCN2018088806-appb-000005
(其中C是大于0小于0.5的常数)。
另一方面,全景视频反映射过程是基于主视点,将二维图像或视频B映射回球面的过程;二维图像或视频B包含分辨率为W I′×H I′的矩形平面I′和分辨率为W II′×H II′矩形平面VI′,矩形平面VI′可进一步划分为四个等腰三角形平面II′,III′,IV′,V′,四个等腰三角形的底和高分别为
Figure PCTCN2018088806-appb-000006
所述反映射方法先将上述平面通过等角投影等方法映射到四棱锥上,然后再将四棱锥映射到球面上。进一步的,过球心O建立右手坐标系,Z轴指向球面经度和纬度为(0°,0°)的方向,Y轴指向球面北极的方向,X轴指向球面经度和纬度为(90°,0°)的方向;再建立一个四棱锥W,其底面中心在Z轴正半轴上,顶点在Z轴负半轴上,底面边分别与X轴和Y轴平行,四棱锥的底面中心D与球心O的连线记作l 1,四棱锥的底面边的中心与球心O的连线记作l 2,l 1和l 2所成角的角度为θ,θ表示了主视点区域的大小,将四棱锥的底面记为面I″,与X轴正半轴相交的侧面记为面II″,与Y轴正半轴相交的侧面记为面III″,与X轴负半轴相交的侧面记为面IV″,与Y轴负半轴相交的侧面记为面V″;当主视点的俯仰角(pitch),航偏角(yaw)和翻滚角(roll)分别为(β 1,β 2,β 3)时对应的四棱锥为上述四棱锥Q绕球心O旋转(β 1,β 2,β 3)后对应的四棱锥Q′,过球心O与四棱锥Q′的四条底边作四个扇形平面,四个扇形平面将球面划为两个区域,其中包含主视点的区域称为主视点区域,其对应棱锥的底面,记为区域I,另一区域称为非主视点区域,过球心O与四棱锥Q′的四条侧边作四个扇形平面,四个平面进一步将非主视点区域划分为四个子区域,四个子区域分别对应棱锥的四个侧面II″,III″,IV″,V″,分别记作区域II、III、IV、V。本发明的反映射方法将二维图像或视频B包含的分辨率为W I′×H I′的矩形平面I′映射到球面的 主视点区域I上,将二维图像或视频B包含的四个等腰三角形平面II′,III′,IV′,V′映射到球面的非主视点区域II、III、IV、V上,所述参数θ,W I′,H I′,W II′,H II′,β 1,β 2,β 3的值包括但不限于从码流中获取。
针对上述全景图像反映射方法,进一步地,将平面图像或视频B映射回球面的方法为对球面上的所有点具体进行如下操作:
第一步:根据球面点的坐标Coordinate以及β 1,β 2,β 3的值判断其位于区域I、II、III、IV、V中的哪一个区域;如果其位于区域I中,则跳到第二步;如果其位于区域II、III、IV、V中,则跳到第五步;
第二步:将球面坐标为Coordinate的点记为A I′,四棱锥Q′底面对边中心的连线记为m 1和m 2,点A I′在m 1和m 2上的投影分别记作B I′和C I′,四棱锥Q′底面中心记为D;根据A I′的坐标Coordinate以及β 1,β 2,β 3的值来计算角B I′OD和角C I′OD的大小;
第三步:根据角B I′OD和角C I′OD的大小来计算平面坐标(X I′,Y I′)的值,X I′和Y I′的值分别与角B I′OD和角C I′OD的大小分别成一次函数关系;
第四步:取矩形平面I′上(X I′,Y I′)处的像素值(或附近像素进行插值),作为球面上坐标为Coordinate的点的像素值;跳过后续步骤;
第五步:将球面坐标为Coordinate的点记为A II′,四棱锥Q′侧面底边中心和顶点的连线记为n 1,过n 1于球心O的平面记为α 3,四棱锥Q′顶点和底面中心的连线记为n 2,点A II′在α 3和n 2上的投影分别记作B II′和C II′,四棱锥Q′底面中心记为D;根据A II′的坐标Coordinate以及β 1,β 2,β 3的值来计算角B II′OD和角A II′C II′B II′的大小;
第六步:根据角B II′OD和角A II′C II′B II′的大小计算出平面坐标(X II′,Y II′)的值(X和Y坐标轴分别垂直和平行于等腰三角形的底边),Y II′的值与角A II′C II′B II′的大小成一次函数关系,角B II′OD的大小与X II′的关系为∠B II′OD=f(X II′),f函数为满足以下条件的任意函数:
θ=f(0)
180°=f(H)
其中H为等腰三角形的高,对于四个等腰三角形平面II′,III′,IV′,V′,其值分别为
Figure PCTCN2018088806-appb-000007
第七步:取三角形平面上(X II′,Y II′)处的像素值(或附近像素进行插值),作为球面上坐标为Coordinate的点的像素值;
对球面上所有的点进行第一步到第七步的操作,由此得到球面的全景图像。
针对上述全景视频反映射过程,进一步的,第六步中,f(X II′)可为:
Figure PCTCN2018088806-appb-000008
(其中C是大于0小于0.5的常数)。
与现有技术相比,本发明的有益效果是:
本发明提供了一种新的全景视频非对称映射方法及相应的反映射方法,通过映射方法将全景图像或视频A所对应的球面映射到二维图像或视频B上;先将球面投影到底面为方形的等腰四棱锥上,再进一步将四棱锥投影到平面上;投影中对主视点的区域使用等角投影并使用较高的采样密度,保证主视点的区域的视频质量较高,对非主视点区域使用较低的采样密度以节省码率。该全景视频非对称映射技术在保证主视点区域视频质量不变的情况下,大大降低视频其余区域的分辨率,有效地节省了虚拟现实视频编码所需的码率。全景视频非对称反映射技术提供了从平面映射回球面的方法,通过此方法可将平面视频映射回球面进行渲染观看。
本发明克服了现有技术的不足,进一步提高非对称映射技术的编码效率,具有以下优点:
(一)本发明中主视点区域球面上的角度与平面坐标成一次函数的关系,即对主视点根据角度等间距的采样,保证主视点区域的采样比较均匀;
(二)本发明中的全景图像映射参数θ和函数f是可调的,即主视点区域的范围以及采样密度的变化速度是可调的;
(三)在合理设置参数θ和函数f(X II)时,能保证主视点区域的质量,又能大大节省码率。如
Figure PCTCN2018088806-appb-000009
时,将分辨率为4096x2048的视频映射成分辨率为平面I′和平面VI′分辨率均为1024x1024的视频,然后将平面I和平面II拼接成分辨率均为2048x1024的视频,相比于已有的非对称映射技术能提高约10%-30%的编码效率;
(四)通过反映射方法,可将本发明中的平面图像B映射回球面进行渲染观看。
附图说明
图1是本发明映射过程中,球面与平面映射关系示意图;
其中,(a)为当主视点的俯仰角(pitch),航偏角(yaw)和翻滚角(roll)分别为(0°,0°,0°)时对应的四棱锥和球面的示意图;(b)为当主视点的俯仰角(pitch),航偏角(yaw)和翻滚角(roll)分别为(β 1,β 2,β 3)时对应的四棱锥和球面的示意图;(c)为将球面通过本发明方法映射到平面后得到的矩形平面I′和矩形平面VI′的示意图;
(a)中:O为球心;Z轴指向球面经度和纬度为(0°,0°)的方向,Y轴指向球面北极的方向,X轴指向球面经度和纬度为(90°,0°)的方向;四棱锥中,底面中心D在Z轴正半轴上,顶点在Z轴负半轴上,底面边分别与X轴和Y轴平行;四棱锥的底面中心D与球心O的连线和四棱锥的底面边的中心与球心O的连线所成角的角度为θ,θ表示主视点区域的大小;
(b)中:四棱锥是将(a)中四棱锥根据俯仰角(pitch),航偏角(yaw)和翻滚角(roll)绕球心O进行旋转后的结果;
(c)中:I′为矩形平面;II′,III′,IV′,V′为四个等腰三角形平面,II′,III′,IV′,V′拼接成矩形平面VI′
图2是本发明中矩形面平I′中坐标为(X I′,Y I′)的点与四棱锥上的点A I′的映射关系示意图;
其中,(a)是矩形面平I′中坐标为(X I′,Y I′)的点的示意图;(b)是四棱锥上的点A I′的示意图;图中,O是球心,D是主视点中心,m 1和m 2分别是四棱锥底面对边中点的连线,点A I′是矩形面平I′中坐标为(X I′,Y I′)的点映射到四棱锥上的对应点,B I′和C I′分别是点A I′在m 1和m 2上的投影。
图3是本发明中三角形平面中坐标为(X II′,Y II′)的点与四棱锥上的点A II′的映射关系示意图;
其中,(a)是三角形平面中坐标为(X II′,Y II′)的点的示意图;(b)是四棱锥上的点A II′的示意图;图中,O是球心,D是主视点中心,n 1是四棱锥侧面底边中心和顶点的连线,n 2是四棱锥顶点和底面中心D的连线,点A II′是三角形平面中坐标为(X II′,Y II′)的点映射到四棱锥上的对应点,B II′和C II′分别是点A II′在n 1和n 2上的投影。
图4是本发明全景图像映射方法实施例效果图。
图5是本发明全景图像反映射方法中计算角BOD和角COD时的示意图;
其中,点B和点C分别是点A在ZOX平面和ZOY平面上投影;D是Z轴正半轴上的点,E是Z轴负半轴上的点。
图6是本发明全景图像反映射方法中球面上位于区域I中的点A I′与矩形平面I′中坐标为(X I′,Y I′)的点的映射关系示意图;
其中,(a)是球面上位于区域I中的点A I′的示意图,O是球心,D是主视点中心,m 1和m 2分别是四棱锥底面对边中点的连线,B I′和C I′分别是点A I′在m 1和m 2上的投影;(b)是矩形平面I′中坐标为(X I′,Y I′)的点的示意图。
图7本发明全景图像反映射方法中球面上位于区域II、III、IV、V中的点A II′与三角形平面中的点(X II′,Y II′)的映射关系示意图;
其中,(a)是球面上位于区域II、III、IV、V中的点A II′的示意图,O是球心,D是主视点中心,n 1是四棱锥侧面底边中心和顶点的连线,n 2是四棱锥顶点和底面中心D的连线,B II′和C II′分别是点A II′在n 1和n 2上的投影;(b)是三角形平面中的点(X II′,Y II′)的示意图。
具体实施方式
下面结合附图,通过实施例进一步描述本发明,但不以任何方式限制本发明的范围。
本发明实施例提供了一种基于主视点的全景图像映射方法,包括全景图像映射方法以及相应的反映射方法,以下分别介绍映射方法以及反映射方法的实施例。
在全景图像映射方法的实施例中,我们将某种映射格式的全景图像A(如经纬图,立方体映射图像等)所对应的球面映射到本发明中设计的全景映射对应的平面图像B,平面图像B包含分辨率为W I′×H I′的矩形平面I′和分辨率为W II′×H II′矩形平面VI′。全景图像映射方法先将球面投影到底面为方形的等腰四棱锥上,再进一步将四棱锥投影到平面上;投影中对主视点的区域使用等角投影并使用较高的采样密度,保证主视点的区域的视频质量较高,对非主视点区域使用较低的采样密度以节省码率。如图1所示,过球心O建立右手坐标系,Z轴指向球面经度和纬度为(0°,0°)的方向,Y轴指向球面北极的方向,X轴指向球面经度和纬度为(90°,0°)的方向;再建立一个四棱锥W,其底面中心在Z轴正半轴上,顶点在Z轴负半轴上,底面边分别与X轴和Y轴平行,四棱锥的底面中心D与球心O的连线记作l 1,四棱锥的底面边的中心与球心O的连线记作l 2,l 1和l 2所成角的角度为θ,θ表示了主视点区域的大小,将四棱锥的底面记为面I″,与X轴正半轴相交的侧面记为面II″,与Y轴正半轴相交的侧 面记为面III″,与X轴负半轴相交的侧面记为面IV″,与Y轴负半轴相交的侧面记为面V″;当主视点的俯仰角(pitch),航偏角(yaw)和翻滚角(roll)分别为(β 1,β 2,β 3)时对应的四棱锥为上述四棱锥Q绕球心O旋转(β 1,β 2,β 3)后对应的四棱锥Q′,过球心O与四棱锥Q′的四条底边作四个扇形平面,四个扇形平面将球面划为两个区域,其中包含主视点的区域称为主视点区域,其对应棱锥的底面,记为区域I,另一区域称为非主视点区域,过球心O与四棱锥Q′的四条侧边作四个扇形平面,四个平面进一步将非主视点区域划分为四个子区域,四个子区域分别对应棱锥的四个侧面II″,III″,IV″,V″,分别记作区域II、III、IV、V。本发明的映射方法将上述的区域I映射到分辨率为W I′×H I′的矩形平面I′上,将区域II、III、IV、V分别映射到四个等腰三角形平面II′,III′,IV′,V′上,四个等腰三角形平面的底和高分别为
Figure PCTCN2018088806-appb-000010
再将四个等腰三角形拼接成分辨率为W II′×H II′矩形平面VI′。所述参数θ,W I′,H I′,W II′,H II′,β 1,β 2,β 3均自行设置。全景图像映射方法具体的流程如下:
第一步:对矩形平面I′中的每个像素点,根据其在平面I′中的坐标(X I′,Y I′),计算其对应到四棱锥Q′底面上的坐标Coordinate′,然后进一步根据透视投影的方法计算其对应到球面上的坐标Coordinate,最后根据球面坐标Coordinate取球面上对应位置的像素值(或附近像素通过插值计算得到相应像素值),作为平面I′中像素点(X I,Y I)的像素值,根据平面坐标(X I′,Y I′)计算球面坐标Coordinate的具体步骤为:
(1.1)如图2所示,平面I′中坐标为(X I′,Y I′)的点对应四棱锥Q′底面上的点为A I′,四棱锥Q′底面对边中心的连线记为m 1和m 2,点A I′在m 1和m 2上的投影分别为B I′和C I′,根据(X I′,Y I′)的值来计算角B I′OD和角C I′OD的大小,其计算公式为:
Figure PCTCN2018088806-appb-000011
Figure PCTCN2018088806-appb-000012
(1.2)根据角B I′OD和角C I′OD的大小以及β 1,β 2,β 3的值求出A I′的坐标Coordinate′,Coordinate′由笛卡尔坐标来表示,其计算公式为:
Figure PCTCN2018088806-appb-000013
Figure PCTCN2018088806-appb-000014
其中T是一个根据俯仰角β 1,航偏角β 2和翻滚角β 3生成的3×3维的旋转矩阵;k是线段OD的长度;
(1.3)根据点A I′的坐标,求出射线
Figure PCTCN2018088806-appb-000015
与球面的交点坐标Coordinate,坐标Coordinate由笛卡尔坐标来表示,其计算公式为:
Figure PCTCN2018088806-appb-000016
其中,R是球的半径;
Figure PCTCN2018088806-appb-000017
表示OA I′线段的模,即OA I′线段的长度,
Figure PCTCN2018088806-appb-000018
除以
Figure PCTCN2018088806-appb-000019
即是对
Figure PCTCN2018088806-appb-000020
进行归一化操作。
需说明,上述步骤(1.2)中k的值未知,但不影响后续计算,因为步骤(1.3)中对
Figure PCTCN2018088806-appb-000021
进行归一化操作时,k会被消除掉。
第二步:对四个等腰三角形平面II′,III′,IV′,V′中的每个像素点,根据其坐标(X II′,Y II′)(X和Y坐标轴分别垂直和平行于等腰三角形的底边),计算其对应到四棱锥Q′侧面上的坐标Coordinate′,然后进一步根据透视投影的方法计算其对应到球面上的坐标Coordinate,最后根据球面坐标Coordinate取球面上对应位置的像素值(或附近像素通过插值计算得到相应像素值),作为像素点(X II′,Y II′)的像素值,根据平面坐标(X II′,Y II′)计算球面坐标Coordinate的具体步骤为:
(2.1)如图3所示,将像素点(X II′,Y II′)对应到四棱锥Q′侧面上的点为A II′,四棱锥Q′侧面底边中心和顶点的连线为n 1,四棱锥Q′顶点和底面中心的连线为n 2,点A II′在n 1和n 2上的投影分别为B II′和C II′,根据(X II′,Y II′)的值来计算角B II′OD和角A II′C II′B II′的大小,其计算公式如下:
∠B II′OD=f(X II′)
Figure PCTCN2018088806-appb-000022
其中,f函数为满足以下条件的任意函数:
θ=f(0)
180°=f(H)
其中H为等腰三角形的高,对于四个等腰三角形平面II′,III′,IV′,V′,其值分别为
Figure PCTCN2018088806-appb-000023
L为等腰三角形的底,对于四个等腰三角形平面II′,III′,IV′,V′,其值分别为W II′,H II′,W II′,H II′
(2.2)根据角B II′OD和角A II′C II′B II′的大小以及β 1,β 2,β 3的值求出A II′的坐标Coordinate′,Coordinate′由笛卡尔坐标来表示,其计算公式为:
Figure PCTCN2018088806-appb-000024
其中T是一个根据俯仰角β 1,航偏角β 2和翻滚角β 3生成的3×3维的旋转矩阵;k是线段C II′B II′的长度;
(2.3)根据点A II′的坐标,求出射线
Figure PCTCN2018088806-appb-000025
与球面的交点坐标Coordinate,坐标Coordinate由笛卡尔坐标来表示,其计算公式为:
Figure PCTCN2018088806-appb-000026
其中,R是球的半径;
Figure PCTCN2018088806-appb-000027
表示OA I′线段的模,即OA I′线段的长度,
Figure PCTCN2018088806-appb-000028
除以
Figure PCTCN2018088806-appb-000029
即是对
Figure PCTCN2018088806-appb-000030
进行归一化操作。
需说明,上述步骤(2.2)中k的值未知,但不影响后续计算,因为步骤(2.3)中对
Figure PCTCN2018088806-appb-000031
进行归一化操作时,k会被消除掉。
第三步:将第二步中得到的四个等腰三角形平面II′,III′,IV′,V′拼接成分辨率为W II′×H II′矩形平面VI′;
至此,全景图像映射方法的实施例的所有步骤完成,实施例展示效果如图4所示。
另一方面,在全景图像反映射方法的实施例中,基于主视点,将二维图像或视频B映射回球面;二维图像或视频B包含分辨率为W I′×H I′的矩形平面I′和分辨率为W II′×H II′矩形平面VI′,矩形平面VI′可进一步划分为四个等腰三角形平面II′,III′,IV′,V′,四个等腰三角形的底和高分别为
Figure PCTCN2018088806-appb-000032
所述反映射方法先将上述平面通过等角投影等方法映射到四棱锥上,然后再将四棱锥映射到球面上。进一步的,过球心O建立右手坐标系,Z轴指向球面经度和纬度为(0°,0°)的方向,Y轴指向球面北极的方向,X轴指向球面经度和纬度为(90°,0°)的方向;再建立一个四棱锥W,其底面中心在Z轴正半轴上,顶点在Z轴负半轴上,底面边分别与X轴和Y轴平行,四棱锥的底面中心D与球心O的连线记作l 1,四棱锥的底面边的中心与球心O的连线记作l 2,l 1和l 2所成角的角度为θ,θ表示了主视点区域的大小,将四棱锥的底面记为面I″,与X轴正半轴相交的侧面记为面II″,与Y轴正半轴相交的侧面记为面III″,与X轴负半轴相交的侧面记为面IV″,与Y轴负半轴相交的侧面记为面V″;当主视点的俯仰角(pitch),航偏角(yaw)和翻滚角(roll)分别为(β 1,β 2,β 3)时对应的四棱锥为上述四棱锥Q绕球心O旋转(β 1,β 2,β 3)后对应的四棱锥Q′,过球心O与四棱锥Q′的四条底边作四个扇形平面,四个扇形平面将球面划为两个区域,其中包含主视点的区域称为主视点区域,其对应棱锥的底面,记为区域I,另一区域称为非主视点区域,过球心O与四棱锥Q′的四条侧边作四个扇形平面,四个平面进一步将非主视点区域划分为四个子区域,四个子区域分别对应棱锥的四个侧面II″,III″,IV″,V″,分别记作区域II、III、IV、V。本发明的反映射方法将二维图像或视频B包含的分辨率为W I′×H I′的矩形平面I′映射到球面的主视点区域I上,将二维图像或视频B包含的四个等腰三角形平面II′,III′,IV′,V′映射到球面的非主视点区域II、III、IV、V上,所述参数θ,W I′,H I′,W II′,H II′,β 1,β 2,β 3的值包括但不限于从码流中获取。全景图像反映射方法具体的流程如下:
第一步:根据球面点的坐标Coordinate以及β 1,β 2,β 3的值判断其位于区域I、II、III、IV、V中的哪一个区域;其具体方法为:
首先根据β 1,β 2,β 3对球面点的坐标Coordinate进行旋转,坐标Coordinate由笛卡尔坐标来表示为(X ,Y ,Z ),旋转后的点记为A,A的坐标为:
(X A,Y A,Z A)=(X ,Y ,Z )×T′
其中,T′是一个根据俯仰角-β 1,航偏角-β 2和翻滚角-β 3生成的3×3维的旋转矩阵;
如图5所示,将点A向ZOX平面和ZOY平面上投影,投影点分别为点B和点C,计算角BOD和角COD的大小:
∠COD=arctan(Y A,Z A)
∠BOD=arctan(X A,Z A)
根据角BOD和角COD判断其所在区域:
Figure PCTCN2018088806-appb-000033
如果不是以上情况,则点在区域II、III、IV、V中,需根据X A,Y A的值进一步判断:
Figure PCTCN2018088806-appb-000034
如果其位于区域I中,则跳到第二步;如果其位于区域II、III、IV、V中,则跳到第五步;
第二步:如图6所示,将球面坐标为Coordinate的点记为A I′,坐标Coordinate由笛卡尔坐标来表示为(X ,Y ,Z ),四棱锥Q′底面对边中心的连线记为m 1和m 2,点A I′在m 1和m 2上的投影分别记作B I′和C I′,四棱锥Q′底面中心记为D;根据A I′的坐标Coordinate以及β 1,β 2,β 3的值来计算角B I′OD和角C I′OD的大小:
Figure PCTCN2018088806-appb-000035
Figure PCTCN2018088806-appb-000036
Figure PCTCN2018088806-appb-000037
其中,T′是一个根据俯仰角-β 1,航偏角-β 2和翻滚角-β 3生成的3×3维的旋转矩阵;
角B I′OD和角C I′OD的计算与第一步中角BOD和角COD的计算方法一样,因此,也可直接使用第一步中角BOD和角COD的值作为角B I′OD和角C I′OD的值;
第三步:根据角B I′OD和角C I′OD的大小来计算平面坐标(X I′,Y I′)的值,其计算公式如下:
Figure PCTCN2018088806-appb-000038
Figure PCTCN2018088806-appb-000039
第四步:取矩形平面I′上(X I′,Y I′)处的像素值(或附近像素进行插值),作为球面上坐标为Coordinate的点的像素值;跳过后续步骤;
第五步:如图7所示,将球面坐标为Coordinate的点记为A II′,坐标Coordinate由笛卡尔坐标来表示为(X ,Y ,Z ),四棱锥Q′侧面底边中心和顶点的连线记为n 1,过n 1于球心O的平面记为α 2,四棱锥Q′顶点和底面中心的连线记为n 2,点A II′在α 3和n 2上的投影分别记作B II′和C II′,四棱锥Q′底面中心记为D;根据A II′的坐标Coordinate以及β 1,β 2,β 3的值来计算角B II′OD和角A II′C II′B II′的大小;
Figure PCTCN2018088806-appb-000040
其中,T′是一个根据俯仰角-β 1,航偏角-β 2和翻滚角-β 3生成的3×3维的旋转矩阵;
Figure PCTCN2018088806-appb-000041
第六步:根据角B II′OD和角A II′C II′B II′的大小计算出平面坐标(X II′,Y II′)的值(X和Y坐标轴分别垂直和平行于等腰三角形的底边),Y II′的值与角A II′C II′B II′的关系为:
Figure PCTCN2018088806-appb-000042
角B II′OD的大小与X II′的关系为∠B II′OD=f(X II′),f函数为满足以下条件的任意函数:
θ=f(0)
180°=f(H)
其中H为等腰三角形的高,对于四个等腰三角形平面II′,III′,IV′,V′,其值分别为
Figure PCTCN2018088806-appb-000043
第七步:取对应三角形平面上(X II′,Y II′)处的像素值(或附近像素进行插值),作为球面上坐标为Coordinate的点的像素值;
至此,完成全景图像反映射方法实施的所有步骤。
需要注意的是,公布实施例的目的在于帮助进一步理解本发明,但是本领域的技术人员可以理解:在不脱离本发明及所附权利要求的精神和范围内,各种替换和修改都是可能的。因此,本发明不应局限于实施例所公开的内容,本发明要求保护的范围以权利要求书界定的范围为准。

Claims (10)

  1. 一种基于主视点的全景视频映射方法,将全景图像或视频A所对应的球面映射到二维图像或视频B上;首先将球面投影到底面为方形的等腰四棱锥上,再进一步将等腰四棱锥投影到平面上;投影中对主视点区域使用等角投影并使用较高的采样密度,使得主视点的区域的视频质量较高;对非主视点区域使用较低的采样密度以节省码率。
  2. 如权利要求1所述全景视频映射方法,其特征是,所述全景视频映射方法包括:
    首先过球心O建立右手坐标系,Z轴指向球面经度和纬度为(0°,0°)的方向,Y轴指向球面北极的方向,X轴指向球面经度和纬度为(90°,0°)的方向;
    再建立一个四棱锥W,底面中心在Z轴正半轴上,顶点在Z轴负半轴上,底面边分别与X轴和Y轴平行,四棱锥的底面中心D与球心O的连线记作l 1,四棱锥的底面边的中心与球心O的连线记作l 2,l 1和l 2所成角的角度为θ,θ表示主视点区域的大小;
    将四棱锥的底面记为面I″,与X轴正半轴相交的侧面记为面II″,与Y轴正半轴相交的侧面记为面III″,与X轴负半轴相交的侧面记为面IV″,与Y轴负半轴相交的侧面记为面V″;当主视点的俯仰角pitch、航偏角yaw和翻滚角roll分别为(β 1,β 2,β 3)时,对应的四棱锥设为四棱锥Q绕球心O旋转(β 1,β 2,β 3)后对应的四棱锥Q′;
    过球心O与四棱锥Q′的四条底边作四个扇形平面,四个扇形平面将球面划为两个区域,其中包含主视点的区域称为主视点区域,其对应棱锥的底面,记为区域I,另一区域称为非主视点区域;
    过球心O与四棱锥Q′的四条侧边作四个扇形平面,四个扇形平面进一步将非主视点区域划分为四个子区域,四个子区域分别对应棱锥的四个侧面II″、III″、IV″、V″,分别记作区域II、III、IV、V;
    将区域I映射到分辨率为W I′×H I′的矩形平面I′上;将区域II、III、IV、V分别映射到四个等腰三角形平面II′,III′,IV′,V′上,四个等腰三角形平面的底和高分别为
    Figure PCTCN2018088806-appb-100001
    Figure PCTCN2018088806-appb-100002
    再将四个等腰三角形拼接成分辨率为W II′×H II′矩形平面VI′;
    矩形平面I′和矩形平面VI′即为映射得到的二维图像或视频B;
    所述参数θ,W I′,H I′,W II′,H II′,β 1,β 2,β 3均可自行设置。
  3. 如权利要求2所述全景视频映射方法,其特征是,所述全景视频映射方法包括如下步骤:
    第一步:对矩形平面I′中的每个像素点,根据其在平面I′中的坐标(X I′,Y I′),计算其对应到四棱锥Q′底面上的坐标Coordinate′;然后进一步根据透视投影的方法计算其对应到球面上的坐标Coordinate;最后根据球面坐标Coordinate取球面上对应位置的像素值,附近像素通过插值计算得到相应像素值,作为平面I′中像素点(X I,Y I)的像素值;
    第二步:对四个等腰三角形平面II′,III′,IV′,V′中的每个像素点,根据其坐标(X II′,Y II′),计算其对应到四棱锥Q′侧面上的坐标Coordinate′,然后进一步根据透视投影的方法计算其对应到球面上的坐标Coordinate,最后根据球面坐标Coordinate取球面上对应位置的像素值,或附近像素通过插值计算得到相应像素值,作为像素点(X II′,Y II′)的像素值;
    第三步:将第二步中得到的四个等腰三角形平面II′,III′,IV′,V′拼接成分辨率为W II′×H II′矩形平面VI′。
  4. 如权利要求3所述全景视频映射方法,其特征是,第一步中,根据平面坐标(X I′,Y I′)计算球面坐标Coordinate的具体步骤为:
    (1.1)将平面I′中坐标为(X I′,Y I′)的点对应四棱锥Q′底面上的点记为A I′,四棱锥Q′底面对边中心的连线记为m 1和m 2,点A I′在m 1和m 2上的投影分别记作B I′和C I′,四棱锥Q′底面中心记为D,A I′的方位可由角B I′OD和角C I′OD来确定;优选地,可根据(X I′,Y I′)的值来计算角B I′OD和角C I′OD的大小,角B I′OD和角C I′OD的大小分别与X I′和Y I′的值成一次函数关系;
    (1.2)根据角B I′OD和角C I′OD的大小以及β 1,β 2,β 3的值求出A I′的坐标Coordinate′;
    (1.3)根据点A I′的坐标,求出射线
    Figure PCTCN2018088806-appb-100003
    与球面的交点坐标Coordinate。
  5. 如权利要求3所述全景视频映射方法,其特征是,第二步中,根据平面坐标(X II′,Y II′)计算球面坐标Coordinate的具体步骤为:
    (2.1)将像素点(X II′,Y II′)对应到四棱锥Q′侧面上的点记为A II′,四棱锥Q′侧面底边中心和顶点的连线记为n 1,四棱锥Q′顶点和底面中心的连线记为n 2,点A II′在n 1和n 2上的投影分别记作B II′和C II′,四棱锥Q′底面中心记为D,A II′的方位可由角B II′OD和角A II′C II′B II′来确定;优选地,可根据(X II′,Y II′)的值来计算角B II′OD和角A II′C II′B II′的大小,角A ′I′C II′B II′的大小与Y II′ 的值成一次函数关系,角B II′OD的大小与X II′的关系为∠B II′OD=f(X II′),f函数为满足以下条件的任意函数:
    θ=f(0)
    180°=f(H)
    其中H为等腰三角形的高,对于四个等腰三角形平面II′,III′,IV′,V′,其值分别为
    Figure PCTCN2018088806-appb-100004
    (2.2)根据角B II′OD和角A II′C II′B II′的大小以及β 1,β 2,β 3的值求出A II′的坐标Coordinate′;
    (2.3)根据点A II′的坐标,求出射线
    Figure PCTCN2018088806-appb-100005
    与球面的交点坐标Coordinate。
  6. 如权利要求5所述全景视频映射方法,其特征是,步骤(2.1)中,f(X II′)为:
    Figure PCTCN2018088806-appb-100006
    其中C是大于0小于0.5的常数。
  7. 如权利要求1所述全景视频映射方法,其特征是,全景图像A的映射格式包括但不限于经纬图、立方体映射图像、多路相机采集的全景视频。
  8. 一种基于主视点的全景视频反映射方法,基于主视点,将二维图像或视频B映射回球面;二维图像或视频B包含分辨率为W I′×H I′的矩形平面I′和分辨率为W II′×H II′矩形平面VI′;矩形平面VI′进一步划分为四个等腰三角形平面II′,III′,IV′,V′,四个等腰三角形的底和高分别为
    Figure PCTCN2018088806-appb-100007
    所述反映射方法先将上述平面通过等角投影等方法映射到四棱锥上,然后再将四棱锥映射到球面上;包括:
    过球心O建立右手坐标系,Z轴指向球面经度和纬度为(0°,0°)的方向,Y轴指向球面北极的方向,X轴指向球面经度和纬度为(90°,0°)的方向;
    再建立一个四棱锥W,其底面中心在Z轴正半轴上,顶点在Z轴负半轴上,底面边分别与X轴和Y轴平行,四棱锥的底面中心D与球心O的连线记作l 1,四棱锥的底面边的中心与球心O的连线记作l 2,l 1和l 2所成角的角度为θ,θ表示了主视点区域的大小,将四棱锥的底面记为面I″,与X轴正半轴相交的侧面记为面II″,与Y轴正半轴相交的侧面记为面III″,与X轴负半轴相交的侧面记为面IV″,与Y轴负半轴相交的侧面记为面V″;
    当主视点的俯仰角pitch,航偏角yaw和翻滚角roll分别为(β 1,β 2,β 3)时,对应的四棱锥设为上述四棱锥Q绕球心O旋转(β 1,β 2,β 3)后对应的四棱锥Q′,过球心O与四棱锥Q′的四条底边作四个扇形平面,四个扇形平面将球面划为两个区域,其中包含主视点的区域称为主视点区域,其对应棱锥的底面,记为区域I,另一区域称为非主视点区域;
    过球心O与四棱锥Q′的四条侧边作四个扇形平面,四个平面进一步将非主视点区域划分为四个子区域,四个子区域分别对应棱锥的四个侧面II″,III″,IV″,V″,分别记作区域II、III、IV、V;
    将二维图像或视频B包含的分辨率为W I′×H I′的矩形平面I′映射到球面的主视点区域I上,将二维图像或视频B包含的四个等腰三角形平面II′,III′,IV′,V′映射到球面的非主视点区域II、III、IV、V上;
    所述参数θ,W I′,H I′,W II′,H II′,β 1,β 2,β 3的值包括但不限于从码流中获取。
  9. 如权利要求8所述全景视频反映射方法,其特征是,将平面图像或视频B映射回球面,针对球面上的所有点进行如下操作:
    第一步:根据球面点的坐标Coordinate以及β 1,β 2,β 3的值判断其位于区域I、II、III、IV、V中的哪一个区域;如果其位于区域I中,则跳到第二步;如果其位于区域II、III、IV、V中,则跳到第五步;
    第二步:将球面坐标为Coordinate的点记为A I′,四棱锥Q′底面对边中心的连线记为m 1和m 2,点A I′在m 1和m 2上的投影分别记作B I′和C I′,四棱锥Q′底面中心记为D;根据A I′的坐标Coordinate以及β 1,β 2,β 3的值来计算角B I′OD和角C I′OD的大小;
    第三步:根据角B I′OD和角C I′OD的大小来计算平面坐标(X I′,Y I′)的值,X I′和Y I′的值分别与角B I′OD和角C I′OD的大小分别成一次函数关系;
    第四步:取矩形平面I′上(X I′,Y I′)处的像素值,或附近像素进行插值,作为球面上坐标为Coordinate的点的像素值;跳过后续步骤;
    第五步:将球面坐标为Coordinate的点记为A II′,四棱锥Q′侧面底边中心和顶点的连线记为n 1,过n 1于球心O的平面记为α 3,四棱锥Q′顶点和底面中心的连线记为n 2,点A II′在α 3和n 2上的投影分别记作B II′和C II′,四棱锥Q′底面中心记为D;根据A II′的坐标Coordinate以及 β 1,β 2,β 3的值来计算角B II′OD和角A II′C II′B II′的大小;
    第六步:根据角B II′OD和角A II′C II′B II′的大小计算出平面坐标(X II′,Y II′)的值,Y II′的值与角A II′C II′B II′的大小成一次函数关系,角B II′OD的大小与X II′的关系为∠B II′OD=f(X II′),f函数为满足以下条件的任意函数:
    θ=f(0)
    180°=f(H)
    其中,H为等腰三角形的高,对于四个等腰三角形平面II′,III′,IV′,V′,其值分别为
    Figure PCTCN2018088806-appb-100008
    第七步:取三角形平面上(X II′,Y II′)处的像素值,或附近像素进行插值,作为球面上坐标为Coordinate的点的像素值;
    对球面上所有的点进行第一步到第七步的操作,由此得到球面的全景图像。
  10. 如权利要求9所述全景视频反映射方法,其特征是,第六步中,f(X II′)可为:
    Figure PCTCN2018088806-appb-100009
    其中,C是大于0小于0.5的常数。
PCT/CN2018/088806 2017-09-26 2018-05-29 基于主视点的全景视频映射方法 Ceased WO2019062177A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US16/650,141 US11301953B2 (en) 2017-09-26 2018-05-29 Main viewpoint-based panoramic video mapping method

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201710882161.1 2017-09-26
CN201710882161.1A CN107622474B (zh) 2017-09-26 2017-09-26 基于主视点的全景视频映射方法

Publications (1)

Publication Number Publication Date
WO2019062177A1 true WO2019062177A1 (zh) 2019-04-04

Family

ID=61090677

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2018/088806 Ceased WO2019062177A1 (zh) 2017-09-26 2018-05-29 基于主视点的全景视频映射方法

Country Status (3)

Country Link
US (1) US11301953B2 (zh)
CN (1) CN107622474B (zh)
WO (1) WO2019062177A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114445279A (zh) * 2021-12-21 2022-05-06 国家电投集团雄安能源有限公司 一种用于双相机拼接管道图像的方法及系统
CN114742704A (zh) * 2022-03-25 2022-07-12 影石创新科技股份有限公司 图像扩增方法和基于图像扩增的模型训练方法
CN115482357A (zh) * 2022-09-27 2022-12-16 北京未尔锐创科技有限公司 一种球面矩形绘制方法及装置

Families Citing this family (18)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2018184528A1 (en) * 2017-04-05 2018-10-11 Mediatek Inc. Method and apparatus for processing projection-based frame with at least one projection face generated using non-uniform mapping
US11057643B2 (en) 2017-03-13 2021-07-06 Mediatek Inc. Method and apparatus for generating and encoding projection-based frame that includes at least one padding region and at least one projection face packed in 360-degree virtual reality projection layout
US11004173B2 (en) 2017-03-13 2021-05-11 Mediatek Inc. Method for processing projection-based frame that includes at least one projection face packed in 360-degree virtual reality projection layout
CN111034191A (zh) 2017-08-18 2020-04-17 联发科技股份有限公司 用于减少基于投影的帧中的伪影的方法和装置
CN107622474B (zh) 2017-09-26 2021-03-30 北京大学深圳研究生院 基于主视点的全景视频映射方法
DE102018006838A1 (de) * 2017-11-03 2019-05-09 Adobe Inc. Dynamisches Mapping von virtuellen und physischen Interaktionen
CN108322727A (zh) * 2018-02-28 2018-07-24 北京搜狐新媒体信息技术有限公司 一种全景视频传输方法及装置
CN108520492B (zh) * 2018-03-16 2022-04-26 中国传媒大学 全景视频映射方法及系统
CN110290285B (zh) * 2018-03-19 2021-01-22 京东方科技集团股份有限公司 图像处理方法、图像处理装置、图像处理系统及介质
CN109792490B (zh) * 2018-06-07 2021-01-15 香港应用科技研究院有限公司 用于流图像压缩的球面视频的改进的伪圆柱映射
US10735765B2 (en) 2018-06-07 2020-08-04 Hong Kong Applied Science and Technology Research Institute Company, Limited Modified pseudo-cylindrical mapping of spherical video using linear interpolation of empty areas for compression of streamed images
CN110956583B (zh) 2018-09-26 2022-05-10 华为技术有限公司 球面图像处理方法、装置及服务器
CN113630622B (zh) * 2021-06-18 2024-04-26 中图云创智能科技(北京)有限公司 全景视频图像处理方法、服务端、目标设备、装置和系统
CN113920277B (zh) * 2021-06-28 2025-01-24 同济大学 一种基于边缘计算存储的Cloud VR数据传输方法和系统
CN115619648A (zh) * 2021-07-14 2023-01-17 华为技术有限公司 全景图像的色调映射方法和装置
CN116309083A (zh) * 2021-12-17 2023-06-23 华为技术有限公司 全景视频的动态范围映射方法及装置
CN115018904B (zh) * 2022-06-02 2023-10-20 如你所视(北京)科技有限公司 全景图像的掩膜生成方法和装置
CN115802039B (zh) * 2023-02-10 2023-06-23 天翼云科技有限公司 一种帧间编码方法、装置、电子设备及计算机可读介质

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104376591A (zh) * 2014-11-18 2015-02-25 北京工业大学 一种手语动态合成光场的生成方法
CN104637061A (zh) * 2015-01-30 2015-05-20 中国科学院自动化研究所 一种二维三维医学图像配准方法
CN104939828A (zh) * 2014-03-28 2015-09-30 阿尔贝特-路德维希斯弗赖堡大学 磁共振成像方法
CN107622474A (zh) * 2017-09-26 2018-01-23 北京大学深圳研究生院 基于主视点的全景视频映射方法

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20030011714A1 (en) * 2001-07-10 2003-01-16 Nevins Robert W. System and method for transmitting program data including immersive content
US20140169699A1 (en) * 2012-09-21 2014-06-19 Tamaggo Inc. Panoramic image viewer
US20140169669A1 (en) * 2012-12-19 2014-06-19 National Chung Cheng University Tone mapping method of high dynamic range image/video
CN105898254B (zh) * 2016-05-17 2018-10-23 北京金字塔虚拟现实科技有限公司 节省带宽的vr全景视频布局方法、装置及展现方法、系统
CN106028115A (zh) * 2016-05-18 2016-10-12 乐视控股(北京)有限公司 视频播放方法和装置
WO2017204491A1 (ko) * 2016-05-26 2017-11-30 엘지전자 주식회사 360 비디오를 전송하는 방법, 360 비디오를 수신하는 방법, 360 비디오 전송 장치, 360 비디오 수신 장치
CN106651764B (zh) * 2016-12-29 2019-10-15 北京奇艺世纪科技有限公司 一种全景图压缩方法及装置
CN106846245B (zh) * 2017-01-17 2019-08-02 北京大学深圳研究生院 基于主视点的全景视频映射方法

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104939828A (zh) * 2014-03-28 2015-09-30 阿尔贝特-路德维希斯弗赖堡大学 磁共振成像方法
CN104376591A (zh) * 2014-11-18 2015-02-25 北京工业大学 一种手语动态合成光场的生成方法
CN104637061A (zh) * 2015-01-30 2015-05-20 中国科学院自动化研究所 一种二维三维医学图像配准方法
CN107622474A (zh) * 2017-09-26 2018-01-23 北京大学深圳研究生院 基于主视点的全景视频映射方法

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
LI, LULU: "Flat Mirror Stitching Measurement Study Based on Structured Light Reflection", CHINA DOCTORAL DISSERTATIONS FULL-TEXT DATABASE, 31 July 2016 (2016-07-31) *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114445279A (zh) * 2021-12-21 2022-05-06 国家电投集团雄安能源有限公司 一种用于双相机拼接管道图像的方法及系统
CN114742704A (zh) * 2022-03-25 2022-07-12 影石创新科技股份有限公司 图像扩增方法和基于图像扩增的模型训练方法
CN115482357A (zh) * 2022-09-27 2022-12-16 北京未尔锐创科技有限公司 一种球面矩形绘制方法及装置

Also Published As

Publication number Publication date
US11301953B2 (en) 2022-04-12
CN107622474B (zh) 2021-03-30
CN107622474A (zh) 2018-01-23
US20210192678A1 (en) 2021-06-24

Similar Documents

Publication Publication Date Title
WO2019062177A1 (zh) 基于主视点的全景视频映射方法
TWI637355B (zh) 紋理貼圖之壓縮方法及其相關圖像資料處理系統與產生360度全景視頻之方法
WO2018157568A1 (zh) 全景图像映射方法
CN106548446B (zh) 一种在球面全景图像上贴图的方法及装置
CN106375748B (zh) 立体虚拟现实全景视图拼接方法、装置及电子设备
WO2018077071A1 (zh) 一种全景图像的生成方法及装置
WO2018068612A1 (zh) 一种全景图像的映射方法、装置和设备
CN106780317B (zh) 全景投影方法及装置
WO2018133380A1 (zh) 基于主视点的全景视频映射方法
WO2017152803A1 (zh) 图像处理方法和设备
CN105137705B (zh) 一种虚拟球幕的创建方法和装置
CN108282694B (zh) 全景视频渲染方法、装置及电子设备
CN106875331B (zh) 一种全景图像的非对称映射方法
CN116363290A (zh) 一种用于大规模场景三维重建的纹理贴图生成方法
CN105809729B (zh) 一种虚拟场景的球形全景渲染方法
WO2020151268A1 (zh) 一种3d小行星动态图的生成方法及便携式终端
WO2017128887A1 (zh) 全景图像的校正3d显示方法和系统及装置
CN1093711C (zh) 全景图象式虚拟现实播放系统和方法
CN108243333A (zh) 全景视频及图片的投影方法和反投影方法
CN118840427B (zh) 基于bim的模型标注和实景点三维定位方法及设备
CN110458960A (zh) 一种基于偏振的彩色物体三维重建方法
WO2022141781A1 (zh) 一种播放全景视频的方法、系统、存储介质及播放设备
CN102855654A (zh) 一种超大规模天气效果渲染方法
CN108765582B (zh) 一种全景图片显示方法及设备
CN112130667A (zh) 一种超高清vr视频的互动方法和系统

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 18860125

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 18860125

Country of ref document: EP

Kind code of ref document: A1