WO2020007094A1 - 一种全景图像滤波方法及装置 - Google Patents

一种全景图像滤波方法及装置 Download PDF

Info

Publication number
WO2020007094A1
WO2020007094A1 PCT/CN2019/083033 CN2019083033W WO2020007094A1 WO 2020007094 A1 WO2020007094 A1 WO 2020007094A1 CN 2019083033 W CN2019083033 W CN 2019083033W WO 2020007094 A1 WO2020007094 A1 WO 2020007094A1
Authority
WO
WIPO (PCT)
Prior art keywords
pixel block
filtered
filtering
image
panoramic
Prior art date
Application number
PCT/CN2019/083033
Other languages
English (en)
French (fr)
Inventor
虞露
皇甫旭昶
Original Assignee
浙江大学
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 浙江大学 filed Critical 浙江大学
Publication of WO2020007094A1 publication Critical patent/WO2020007094A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00Image analysis
    • G06T7/10Segmentation; Edge detection
    • G06T7/11Region-based segmentation

Definitions

  • the present invention relates to the field of panoramic image processing, and in particular, to a method and a device for filtering pixel blocks of a panoramic image.
  • the panoramic image refers to all the scenes around an observation point in space. It consists of all the light that can be received by this observation point.
  • the spherical surface can describe all the scenes around the observation point. Because spherical images are difficult to store and encode, the spherical image of a panoramic image is expanded into a planar image by defining a specific projection expansion method.
  • the planar image is called a panoramic projection expansion image.
  • the projection expansion method generally consists of a certain panoramic image.
  • the image expansion format information is determined.
  • Cube is one of the most common unfolding formats of panoramic image projection.
  • the center projection method is used to project a sphere onto six faces of its corresponding external cube.
  • the eight vertices of the cube are A, B, C, D, E, F, G, H, and then expand the cube, with the left, front, and right sides arranged horizontally, and the top, back, and bottom arranged vertically. Because the left, front, and right sides show the outer surface of the cube, and the top, back, and bottom of the cube show the inner surface of the cube in Figure 1, the top, back, and bottom need to be arranged vertically. Flip left and right to represent the outer surface of the cube.
  • Figure 2 As shown in Figure 2, turn the top, back, and bottom left and right and then rotate 90 ° clockwise to compactly arrange the 6 faces of the cube into a 3x2 rectangular image.
  • Figure 2 only shows An arrangement of the 6 faces of the cube is obtained, and the 6 faces of the cube may not be arranged in this way.
  • the area close to the boundary of the six faces of the cube is called the area of the boundary of the cube.
  • the area of the original boundary adjacent to the spherical surface may not be adjacent in the panoramic projection expanded image.
  • the surface boundary areas 1 and 1 ' have a common boundary AE on the cube, and the boundary areas 1 and 1' are not adjacent on the panoramic projection expanded image.
  • the gray areas represented by the numbers a and a 'in Fig. 3 Represents a pair of surface boundary regions that are adjacent on a spherical surface but not adjacent on a panoramic projection expanded image, where a can be any of 1, 2, 3, 4, 5, 6, 7, and 8.
  • the white area in FIG. 3 represents a non-plane boundary area on the panoramic projection expanded image
  • the black area represents a surface boundary area that is adjacent on the spherical surface and is still adjacent on the panoramic projection expanded image.
  • the traditional deblocking filtering method is two-dimensional separated filtering.
  • the boundary direction of the block and the filtering processing direction are perpendicular to each other, that is, the horizontal boundary filtering processing is performed on the vertical boundary, and the first-dimensional filtering processing direction and the second-dimensional filtering processing direction are also perpendicular to each other.
  • the boundary direction between the pixel blocks A and B is a vertical direction, so firstly perform a first-dimensional filtering process on the vertical boundary between the pixel blocks A and B, that is, first perform horizontal filtering.
  • the pixel blocks A and C are horizontal boundaries, and the results of the first-dimensional filtering process are subjected to the second-dimensional filtering process, that is, vertical filtering is performed.
  • HEVC High Efficiency Video Coding
  • H.264 when performing a deblocking filtering operation on a pixel block boundary, first perform horizontal filtering on the vertical boundary and then perform vertical filtering on the horizontal boundary, that is, first horizontal filtering. Then vertical filtering.
  • encoding or decoding order is from top to bottom from left to right, as shown in Figure 5.
  • pixel block A is first encoded or decoded, and then Pixel block B is encoded or decoded, pixel block C is encoded or decoded, and pixel block D is finally encoded or decoded. Therefore, in FIG. 3, among the pair of surface boundary regions a and a ′ which are adjacent on the spherical surface but not adjacent on the panoramic projection unfolded image, a ′ is a surface boundary region of the later encoding or decoding process, where a takes 1, Any of 2, 3, 4, 5, 6, 7, and 8.
  • the filtering technology in the image codec performs filtering operations on adjacent pixel blocks in the coded image, and does not perform filtering on the pixel blocks on both sides of the spherical segmentation line of the panoramic image.
  • special filtering methods need to be designed for the pixel blocks on both sides of the spherical segmentation line of the panoramic video. Due to the misalignment of the pixel blocks on both sides of the spherical segmentation line of the panoramic video, This phenomenon requires the design of corresponding alignment methods.
  • the object of the present invention is to solve the discontinuity at the boundary between the area boundary areas adjacent to each other on the panoramic spherical surface but not adjacent to each other on the projected expanded image.
  • the direction of the first and second dimension filtering processing of the surface boundary region is determined by specifying a specific surface boundary among the surface boundary region pairs to ensure that matching can be achieved when the codec-end-to-surface boundary region pair is filtered.
  • a first object of the present invention is to provide a panoramic image filtering method, including:
  • the method of the second two-dimensional filtering processing direction is consistent with determining the directions of the first-dimensional and second-dimensional filtering processing on the panoramic projection expanded image of the non-plane boundary area;
  • a first-dimensional and a second-dimensional filtering process are sequentially performed on the pair of surface boundary regions.
  • the specific surface boundary region is a surface boundary region in the pair of surface boundary regions that is later encoded or decoded on the panoramic projection expanded image.
  • the filtered pixels of the common sub-region are the surface boundary regions of the two surface boundary regions that are later encoded or decoded on the panoramic unfolded projection image.
  • the corresponding sub-region filtered pixels are determined.
  • a second object of the present invention is to provide a panoramic image filtering device, including the following modules:
  • the position derivation module for the boundary area of the surface to be filtered The input of this module is the format information of the panoramic projection expanded image. Based on the format information, it is determined that the panoramic projection expanded image is adjacent to the panoramic spherical surface, but is different from the projection expanded image. The position of the adjacent surface boundary region pairs, the output of this module is the position of the pair of surface boundary region pairs to be filtered in the panoramic projection expanded image;
  • Filter processing direction determination module This module inputs the pair of surface boundary regions to be filtered and their position in the panoramic projection expansion image, and determines the direction of a specific surface boundary region in the panoramic projection expansion image in the surface boundary region pair to determine The direction of the first- and second-dimensional filtering processes in the area of the surface boundary region pair, the method for determining the direction of the first- and second-dimensional filtering process with the area-region boundary pair, and the non-plane-boundary region determining the first and second dimensions on the panoramic projection expanded image
  • the directions of the two-dimensional filtering process are the same; the module outputs the directions of the first-dimensional and second-dimensional filtering processes of the boundary region pair of the surface to be filtered;
  • Filter processing module inputs the first-dimensional and second-dimensional filtering processing directions of the surface boundary region pair to be filtered, and sequentially performs the first-dimensional and second-dimensional filtering processing on the surface boundary region pair.
  • a specific surface boundary region in the filtering processing direction determination module is a surface boundary region that is later encoded or decoded on the panoramic projection expanded image in the pair of surface boundary regions.
  • the filtering processing module for the common sub-regions that belong to the two surface boundary regions, the filtered pixels in the common sub-region are encoded later on the panoramic unfolded projection image in the two surface boundary regions.
  • the sub-region corresponding to the area boundary region of the decoding / decoding process is determined by filtering the pixels.
  • a third object of the present invention is to provide a method for filtering a pixel block of a panoramic image, comprising: according to projection format information corresponding to an image of a pixel block to be filtered immediately adjacent to a side of a spherical segmentation line of the panoramic image; The position of the nominal point along the direction of the spherical dividing line is expanded and contracted to determine the position of the nominal point corresponding to the corresponding nominal point on the other side of the spherical dividing line; the corresponding nominal point is used to determine the target point.
  • the corresponding pixel block of the filtered pixel block immediately adjacent to the other side of the spherical dividing line; filtering the pixel block to be filtered and its corresponding pixel block to obtain a filtered result of the pixel block to be filtered.
  • the method for determining the position of the corresponding nominal point in the corresponding pixel block of the pixel block to be filtered is one of the following cases:
  • the method for determining the positions of the remaining non-corresponding nominal points in the corresponding pixel block of the pixel block to be filtered is: determining the corresponding pixels at the entire pixel interval based on the corresponding nominal points in the corresponding pixel block. The position of the non-corresponding nominal point in the block.
  • the filtering method further includes obtaining coding information of the pixel block to be filtered and its corresponding pixel block, and determining a filtering method according to the coding information of the pixel block to be filtered and its corresponding pixel block.
  • the method for acquiring coding information corresponding to the pixel block to be filtered is one of the following cases:
  • the encoding information comes from the encoding unit where the corresponding nominal point in the corresponding pixel block is located;
  • the coding information is from a coding unit containing the largest number of pixels in the corresponding pixel block.
  • a fourth object of the present invention is to provide a method for filtering a panoramic image pixel block, including:
  • the corresponding pixel of the pixel block to be filtered on the other side of the spherical segmentation line is determined.
  • the pixel block to be filtered and its corresponding pixel block are filtered to obtain the filtered result of the pixel block to be filtered.
  • a fifth object of the present invention is to provide a panoramic image pixel block filtering device, which includes the following modules:
  • Corresponding nominal point position derivation module The input of this module is the position information of the nominal point in the pixel block to be filtered immediately to the side of the spherical segmentation line of the panoramic image and the projection format information corresponding to the image where the pixel block to be filtered is located; the output is The position of the nominal point in the panoramic image corresponding to the nominal point; according to the projection format information corresponding to the image of the pixel block to be filtered on the side of the spherical segmentation line of the panoramic image, the nominal point in the pixel block to be filtered is Performing a telescoping transformation on a position in the direction of the spherical dividing line to determine a position of the nominal point corresponding to a corresponding nominal point on the other side of the spherical dividing line;
  • Corresponding pixel block position derivation module The module input is the position of the corresponding nominal point, and the output is the position of the corresponding pixel block in the panoramic image of the pixel block to be filtered; this module is used to determine the corresponding pixel block according to the position of the corresponding nominal point. The location of the remaining non-corresponding nominal points;
  • Pixel filtering module The input of this module is the position of the pixel block to be filtered and the position of the corresponding pixel block, and the output is the filtered pixel of the pixel block to be filtered; this module is used to filter the pixel block to be filtered and its corresponding pixel block, The filtered result of the pixel block to be filtered is obtained.
  • the corresponding nominal point position derivation module further includes a corresponding nominal point position rounding module, and the rounding method of the corresponding nominal point position in the corresponding nominal point position rounding module is one of the following cases:
  • the corresponding pixel block position derivation module further includes a non-corresponding nominal point position derivation module.
  • the method for determining the non-nominal point position in the non-corresponding nominal point position derivation module is: The nominal point is used as a reference, and the position of the non-corresponding nominal point is determined at the whole pixel interval.
  • the pixel filtering module further includes the following sub-modules:
  • a coding information acquisition module configured to acquire coding information of the pixel block to be filtered and its corresponding pixel block
  • a filtering method determining module configured to determine a filtering method according to the coding information and pixel characteristics of the pixel block to be filtered and its corresponding pixel block.
  • the method for acquiring coding information corresponding to the pixel block to be filtered is one of the following cases:
  • the encoding information comes from the encoding unit where the corresponding nominal point in the corresponding pixel block is located;
  • the coding information is from a coding unit containing the largest number of pixels in the corresponding pixel block.
  • Corresponding pixel block position derivation module The module inputs the position information of the pixel block to be filtered immediately adjacent to the spherical segmentation line side of the panoramic image and the projection format information corresponding to the image where the pixel block to be filtered is located, and outputs the corresponding pixel block to be filtered.
  • a coding information acquisition module configured to obtain coding information of the pixel block to be filtered and its corresponding pixel block, wherein the motion information in the coding information of the pixel block to be filtered and its corresponding pixel block is always considered to be the same;
  • a pixel filtering module is configured to filter the pixel block to be filtered and its corresponding pixel block to obtain a filtered result of the pixel block to be filtered.
  • a sixth object of the present invention is to provide a panoramic image pixel block filtering device, which includes the following modules:
  • Corresponding pixel block position derivation module The module inputs the position information of the pixel block to be filtered immediately adjacent to the spherical segmentation line side of the panoramic image and the projection format information corresponding to the image where the pixel block to be filtered is located, and outputs the corresponding pixel block to be filtered.
  • a filtering method determining module is used to determine a filtering method.
  • the input is the position information of the pixel block to be filtered and its corresponding pixel block.
  • the filtering method is determined based on the coding information of the pixel block to be filtered and its corresponding pixel block.
  • the output is determined by the module. Filtering method.
  • the determining filtering method is one of the following:
  • a pixel filtering module is configured to filter the pixel block to be filtered and its corresponding pixel block to obtain a filtered result of the pixel block to be filtered.
  • the beneficial effects of the present invention can reduce the coding artifacts of the panoramic image and the discontinuity of the surface boundary, improve the subjective quality of the panoramic image after encoding, and at the same time specify the directions of the first-dimensional and second-dimensional filtering processing to ensure that it is in the opposite boundary area When filtering is performed, matching at the encoding and decoding ends can be achieved.
  • Figure 1 maps the spherical surface to the circumscribed cube of the sphere through perspective projection, and expands the six faces of the circumscribed cube;
  • Figure 2 shows that the six sides of the cube are expanded into a 3x2 format by rotating and compacting
  • FIG. 3 shows the distribution of the boundary area and the non-plane boundary area on the panoramic projection expansion image
  • FIG. 5 is a coding or decoding sequence of pixel blocks in an image in a conventional codec technology
  • FIG. 6 is a pair of surface boundary regions adjacent on a spherical surface but not adjacent on a panoramic projection unfolded image in an embodiment of the present invention
  • FIG. 7 is a block of pixels to be filtered in a pair of surface boundary regions that are adjacent on a spherical surface but not adjacent on a panoramic projection expanded image in an embodiment of the present invention
  • FIG. 8 is a filtering direction and sequence of pixel blocks to be filtered in pair 2 and 2 ', which are adjacent to a spherical surface but not adjacent to a panoramic projection spread image in an embodiment of the present invention
  • FIG. 9 is an alignment manner of the to-be-filtered pixel blocks in pairs 1 and 1 'on a spherical surface but not adjacent on a panoramic projection expanded image according to an embodiment of the present invention
  • FIG. 10 is an alignment manner of pixel blocks to be filtered in pairs 3 and 3 'that are adjacent on a spherical surface but not adjacent on a panoramic projection expanded image according to an embodiment of the present invention
  • FIG. 11 is a block of pixels to be filtered in the pair of surface boundary regions 4 and 4 'that are adjacent on a spherical surface but not adjacent on a panoramic projection expanded image in an embodiment of the present invention
  • FIG. 12 is a filtering direction and sequence of pixel blocks to be filtered in pairs 4 and 4 ', which are adjacent to a spherical surface but not adjacent to a panoramic projection expanded image according to an embodiment of the present invention
  • FIG. 13 is a block of pixels to be filtered in the pair of surface boundary regions 5 and 5 'that are adjacent on the spherical surface but not adjacent on the panoramic projection expanded image in an embodiment of the present invention
  • FIG. 15 is a block of pixels to be filtered in the pair of surface boundary regions 7 and 7 'that are adjacent on the spherical surface but not adjacent on the panoramic projection expanded image in an embodiment of the present invention
  • 16 is a filtering direction and sequence of pixel blocks to be filtered in pair 7 and 7 'which are adjacent to a spherical surface but not adjacent to a panoramic projection unfolded image in an embodiment of the present invention
  • 17 is a device diagram in an embodiment of the present invention.
  • FIG. 18 is a schematic diagram of a common subregion that belongs to the boundary region of two faces in one embodiment of the present invention.
  • FIG. 19 is a flowchart of filtering processing performed on a common sub-region.
  • FIG. 20 is a schematic diagram of the positional relationship after the six faces of the hexahedron are unfolded and compactly arranged;
  • 21 is a schematic diagram of positions of a pixel block to be filtered and its corresponding pixel block in a panoramic image expanded image according to an embodiment of the present invention
  • 22 is a schematic diagram of a nominal point position in a pixel block to be filtered according to an embodiment of the present invention
  • FIG. 23 is a schematic diagram of a position of a corresponding nominal point in a corresponding pixel block according to an embodiment of the present invention.
  • FIG. 24 is a schematic diagram of a position of a corresponding nominal point in a corresponding pixel block according to an embodiment of the present invention.
  • FIG. 25 is a schematic diagram illustrating a position of a nominal point in a pixel block to be filtered according to an embodiment of the present invention
  • 26 is a schematic diagram of a position of a corresponding nominal point in a corresponding pixel block and the coding unit of the corresponding block according to an embodiment of the present invention
  • FIG. 27 is a schematic diagram of positions of a pixel block to be filtered and its corresponding pixel block in a panoramic image expanded into a hexahedral image according to an embodiment of the present invention
  • FIG. 28 is a schematic diagram of the position of a pixel block to be filtered and its corresponding pixel block in a longitude and latitude image in an embodiment of the present invention
  • FIG. 29 is a schematic diagram of a panoramic image pixel block filtering device according to an embodiment of the present invention.
  • FIG. 30 is a schematic diagram of a panoramic image pixel block filtering device according to an embodiment of the present invention.
  • FIG. 31 is a schematic diagram of sub-modules in a corresponding nominal point position deriving module according to an embodiment of the present invention.
  • FIG. 32 is a schematic diagram of a sub-module in a corresponding pixel block position deriving module according to an embodiment of the present invention.
  • FIG. 33 is a schematic diagram of a sub-module in a pixel filtering module according to an embodiment of the present invention.
  • 34 is a schematic diagram of an artifact in a panoramic image viewport caused by encoding
  • Figure 35 is a filtered panoramic image viewport
  • FIG. 36 is a flow chart of boundary strength determination in an embodiment of the present invention.
  • Figure 37 shows the decision flow of boundary strength in HEVC deblocking filtering.
  • An embodiment of the present invention provides a panoramic image filtering method.
  • the method of the second two-dimensional filtering processing direction is consistent with determining the directions of the first-dimensional and second-dimensional filtering processing on the panoramic projection expanded image of the non-plane boundary area;
  • a first-dimensional and a second-dimensional filtering process are sequentially performed on the pair of surface boundary regions.
  • the specific area boundary area is the area area of the area boundary area centered on the panoramic projection expanded image after being encoded or decoded.
  • the panoramic projection expanded image in the cube format as shown in FIG. 6 is arranged in a 3x2 rectangular image, so it can be easily determined that the surface boundary areas a and a ′ on the panoramic projection expanded image shown in FIG. 6 are in the panoramic image.
  • the pixel block b in the surface boundary region 2 and the pixel block b ′ in the surface boundary region 2 ′ are pixel blocks that are adjacent on the panoramic spherical surface but are not adjacent on the panoramic projection expanded image, so Pixel blocks b and b 'need to be deblocked to smooth discontinuities between the pixel blocks.
  • the surface boundary region 2 ′ is a region subjected to later encoding or decoding processing, and the direction of the surface boundary region 2 ′ on the panoramic projection expanded image is vertical.
  • the direction of the first-dimensional filtering process in the non-plane boundary region is set to the horizontal direction
  • the direction of the second-dimensional filtering process is set to the vertical direction.
  • the present invention is not limited to this setting. For example, it can also be set to the first-dimensional filtering process.
  • the direction of is the vertical direction
  • the direction of the second-dimensional filtering process is the horizontal direction.
  • first-dimensional filtering process direction of the face boundary region pair 2 and 2 ' is also horizontal, and the pixel blocks in the opposite boundary region pair 2 and 2' are in order.
  • First and second dimension filtering processes are performed.
  • FIG. 7 there are pixel blocks d and e in the surface boundary region 2 and pixel blocks d 'and e' in the surface boundary region 2 '.
  • the pixel blocks d, e, d', and e ' are filtered.
  • the pixel blocks d and e are rotated 90 ° counterclockwise, and then aligned with the pixel blocks d ′ and e ′, and then the pixel blocks are sequentially processed in the first and second dimensions.
  • FIG. 7 there are pixel blocks d and e in the surface boundary region 2 and pixel blocks d 'and e' in the surface boundary region 2 '.
  • the pixel blocks d and d ′, e and e ′ are firstly subjected to a deblocking filtering process in the horizontal direction to obtain the filtered result, and then the pixel blocks d ′ and e ′, d and e are vertically adjusted.
  • Deblocking filtering in straight direction According to the deblocking filtering process adopted in this embodiment, the order of the first-dimensional and second-dimensional filtering processes of the pixel blocks in the surface boundary region 2 is different from the order of the first-dimensional and second-dimensional filtering processes of the pixel blocks in the non-surface boundary region.
  • the filtering direction is also determined according to the method described in this embodiment to filter the boundaries of the surface boundary region pairs.
  • the direction of the surface boundary region 1' on the panoramic projection unfolded image is horizontal, and the filtering order of the pixel blocks in 1 and 1 'by the surface boundary region is the filtering of the vertical boundary first. Then the horizontal boundary filtering is performed.
  • the pixel blocks in the surface boundary region 1 need not be subjected to any rotation or transposition to directly perform deblocking filtering processing on the pixel blocks in the surface boundary region 1 ′, as shown in FIG. 9.
  • the order of the first-dimensional and second-dimensional filtering processes of the pixels in 1 is not changed; among the surface boundary regions 3 and 3 ', the direction of 3' on the panoramic projection expanded image is the horizontal direction. Then the order of filtering the pixel blocks in 3 and 3 'in the boundary area of the area is first the filtering of the vertical boundary and then the filtering of the horizontal boundary. It is necessary to rotate the pixel blocks in the surface boundary area 3 by 180 ° and the surface boundary area 3' The deblocking filtering process is performed. As shown in FIG. 10, the order of the first-dimensional and second-dimensional filtering processes of the pixels in the surface boundary region pair 3 is not changed.
  • the pixel blocks in the pair of surface boundary regions that are adjacent on the spherical surface but not adjacent on the panoramic projection unfolded image in FIG. 6 are also processed according to the method described in this embodiment.
  • An embodiment of the present invention provides a panoramic image filtering method.
  • the method of the second two-dimensional filtering processing direction is consistent with determining the directions of the first-dimensional and second-dimensional filtering processing on the panoramic projection expanded image of the non-plane boundary area;
  • a first-dimensional and a second-dimensional filtering process are sequentially performed on the pair of surface boundary regions.
  • the specific surface boundary region is a surface boundary region that is firstly encoded or decoded on the panoramic projection unfolded image in the area of the surface boundary region pairing;
  • the region of boundary region 4 is a region that is encoded or decoded first, and the direction of the region of boundary region 4 on the panoramic projection expanded image is vertical.
  • it is set
  • the direction of the first-dimensional filtering process in the non-face boundary region is horizontal, and the direction of the second-dimensional filtering process is vertical.
  • the present invention is not limited to this setting.
  • the direction of the first-dimensional filtering process may be set as In the vertical direction
  • the direction of the second-dimensional filtering process is the horizontal direction.
  • first-dimensional filtering process direction of the face boundary region pairs 4 and 4 ' is also horizontal, and the pixel blocks in the opposite boundary region pair 4 and 4' are in order.
  • First and second dimension filtering processes are performed.
  • FIG. 11 there are pixel blocks b and c in the surface boundary region 4 and pixel blocks b 'and c' in the surface boundary region 4 '.
  • the pixel blocks b, c, b', and c ' are filtered.
  • the pixel blocks b ′ and c ′ are rotated counterclockwise by 90 ° and aligned with the pixel blocks b and c, and then the pixel blocks are sequentially processed in the first and second dimensions.
  • FIG. 11 there are pixel blocks b and c in the surface boundary region 4 and pixel blocks b 'and c' in the surface boundary region 4 '.
  • the pixel blocks b ′ and b, c, and c ′ are subjected to a horizontal deblocking filtering process to obtain a result of the filtering process, and then the pixel blocks b ′ and c ′, b, and c are vertically adjusted.
  • Deblocking filtering in straight direction According to the deblocking filtering process adopted in this embodiment, the order of the first-dimensional and second-dimensional filtering processes of the pixel blocks in the surface boundary region 4 is different from the order of the first-dimensional and second-dimensional filtering processes of the pixel blocks in the non-plane boundary region.
  • the pixel blocks in the pair of surface boundary regions that are adjacent on the spherical surface but not adjacent on the panoramic projection unfolded image in FIG. 6 are also processed according to the method described in this embodiment.
  • An embodiment of the present invention provides a panoramic image filtering method.
  • the method of the second two-dimensional filtering processing direction is consistent with determining the directions of the first-dimensional and second-dimensional filtering processing on the panoramic projection expanded image of the non-plane boundary area;
  • a first-dimensional and a second-dimensional filtering process are sequentially performed on the pair of surface boundary regions.
  • the specific plane boundary region is a plane boundary region in the horizontal direction of the plane boundary region pair.
  • the plane boundary region 5 and the plane boundary region 5' have different directions on the panoramic projection expanded image.
  • the plane boundary region 5 ' is a plane boundary region in the horizontal direction.
  • the direction of the first-dimensional filtering process in the non-plane boundary region is set to the vertical direction
  • the direction of the second-dimensional filtering process is set to the horizontal direction. It is not limited to this setting, for example, the direction of the first-dimensional filtering process may be set to a horizontal direction, and the direction of the second-dimensional filtering process may be set to a vertical direction.
  • the first-dimensional filtering process direction of the face boundary region pair 5 and 5 ' is also the vertical direction, and the pixels in the opposite boundary region pair 5 and 5' are in the vertical direction.
  • the block performs first-dimensional and second-dimensional filtering processes in this order.
  • FIG. 13 there are pixel blocks b and c in the surface boundary region 5 and pixel blocks b 'and c' in the surface boundary region 5 '.
  • the pixel blocks b, c, b', and c ' are filtered.
  • the pixel blocks b and c are rotated 90 ° clockwise, and then aligned with the pixel blocks b ′ and c ′, and then the pixel blocks are sequentially processed in the first dimension and the second dimension.
  • FIG. 13 there are pixel blocks b and c in the surface boundary region 5 and b 'and c' in the surface boundary region 5 '.
  • the pixel blocks b ′ and b, c ′ and c are subjected to a vertical deblocking filtering process to obtain a filtering process result, and then the pixel blocks b and c, b ′ and c ′ are processed.
  • Deblocking filtering in the horizontal direction According to the deblocking filtering process used in this embodiment, the order of the first-dimensional and second-dimensional filtering processes of the pixel blocks in the surface boundary region 5 is the same as the order of the first-dimensional and second-dimensional filtering processes of the pixel blocks in the non-plane boundary region.
  • the pixel blocks in the pair of surface boundary regions that are adjacent on the spherical surface but not adjacent on the panoramic projection unfolded image in FIG. 6 are also processed according to the method described in this embodiment.
  • An embodiment of the present invention provides a panoramic image filtering method.
  • the method of the second two-dimensional filtering processing direction is consistent with determining the directions of the first-dimensional and second-dimensional filtering processing on the panoramic projection expanded image of the non-plane boundary area;
  • a first-dimensional and a second-dimensional filtering process are sequentially performed on the pair of surface boundary regions.
  • the specific plane boundary region is a plane boundary region in the vertical direction of the plane boundary region pair.
  • the plane boundary region 7 and the plane boundary region 7' have different directions on the panoramic projection expanded image.
  • the plane boundary region 7 ' is a plane boundary region in the vertical direction.
  • the direction of the first-dimensional filtering process in the non-plane boundary region is set to the horizontal direction
  • the direction of the second-dimensional filtering process is set to the vertical direction. It is not limited to this setting.
  • the direction of the first-dimensional filtering process may be set to a vertical direction
  • the direction of the second-dimensional filtering process may be set to a horizontal direction.
  • first-dimensional filtering process direction of the face boundary region pair 7 and 5 ' is also horizontal, and the pixel blocks in the opposite boundary region pair 7 and 7' are in order.
  • First and second dimension filtering processes are performed.
  • FIG. 15 there are pixel blocks b and c in the surface boundary region 7 and pixel blocks b 'and c' in the surface boundary region 7 '.
  • the pixel blocks b, c, b', and c ' are filtered.
  • the pixel blocks b and c are rotated 90 ° clockwise, and then aligned with the pixel blocks b ′ and c ′, and then the pixel blocks are sequentially processed in the first dimension and the second dimension.
  • FIG. 15 there are pixel blocks b and c in the surface boundary region 7 and b 'and c' in the surface boundary region 7 '.
  • the pixel blocks b ′ and b, c ′, and c are subjected to a horizontal deblocking filtering process to obtain a result of the filtering process, and then the pixel blocks b and c, b ′, and c ′ are vertically adjusted.
  • Deblocking filtering in straight direction According to the deblocking filtering process used in this embodiment, the order of the first-dimensional and second-dimensional filtering processes of the pixel blocks in the surface boundary region 7 is different from the order of the first-dimensional and second-dimensional filtering processes of the pixel blocks in the non-plane boundary region.
  • the pixel blocks in the pair of surface boundary regions that are adjacent on the spherical surface but not adjacent on the panoramic projection unfolded image in FIG. 6 are also processed according to the method described in this embodiment.
  • An embodiment of the present invention provides a panoramic image filtering device.
  • the position derivation module for the boundary area of the surface to be filtered The input of this module is the format information of the panoramic projection expanded image. Based on the format information, it is determined that the panoramic projection expanded image is adjacent to the panoramic spherical surface, but is different from the projection expanded image The position of the adjacent surface boundary region pairs, the output of this module is the position of the pair of surface boundary region pairs to be filtered in the panoramic projection expanded image;
  • Filter processing direction determination module This module inputs the pair of surface boundary regions to be filtered and their position in the panoramic projection expansion image, and determines the direction of a specific surface boundary region in the panoramic projection expansion image in the surface boundary region pair to determine The direction of the first- and second-dimensional filtering processes in the area of the surface boundary region pair, the method for determining the direction of the first- and second-dimensional filtering process with the area-region boundary pair, and the non-plane-boundary region determining the first and second dimensions on the panoramic projection expanded image
  • the directions of the two-dimensional filtering process are the same; the module outputs the directions of the first-dimensional and second-dimensional filtering processes of the boundary region pair of the surface to be filtered;
  • Filter processing module inputs the first-dimensional and second-dimensional filtering processing directions of the surface boundary region pair to be filtered, and sequentially performs the first-dimensional and second-dimensional filtering processing on the surface boundary region pair.
  • FIG. 17 is a schematic diagram of a panoramic image filtering device in this embodiment.
  • the format information of the input panoramic projection expansion image determines the expansion method of the panoramic projection expansion image, such as cube format, EAC, PAU, etc.
  • This embodiment uses the cube format as an example.
  • the panoramic projection expanded image is projected onto six faces of the cube and the six faces are arranged into a 3 ⁇ 2 rectangular image.
  • the area corresponding to the same number shown in FIG. 6 is the boundary area pair of the surface to be filtered.
  • the filtering processing direction determination module it is necessary to specify one of the surface boundary region pairs, and specify the first dimension and the second dimension of the surface boundary region pair according to the direction of the surface boundary region in the panoramic projection expanded image. Dimensional filtering processing direction.
  • the specified polygon boundary area is one of the following 4 cases:
  • the specified surface boundary region is a surface boundary region in which the surface boundary region is firstly encoded or decoded on the panoramic projection expanded image;
  • the specified surface boundary region is a surface boundary region in which the surface boundary region is centered on the panoramic projection expanded image after being encoded or decoded;
  • the specified surface boundary region is a horizontal surface boundary region in the surface boundary region pair
  • the specified face boundary region is a face boundary region in the vertical direction of the face boundary region pair;
  • the filtering processing direction determining module specifies the first-dimensional and second-dimensional filtering processing directions of the pair of surface boundary regions according to the direction of the specified surface boundary region.
  • the filtering processing module sequentially performs first-dimensional and second-dimensional filtering processing on the pair of surface boundary regions.
  • the filtering processing module for the common sub-regions that belong to the boundary regions of two planes, the filtered pixels in the common sub-region are compared by the two plane-boundary regions according to the encoding and decoding order on the panoramic expanded projection image.
  • the post-encoding / decoding process determines the filtered pixels of the sub-region corresponding to the area boundary region.
  • the pixel blocks in the pair of surface boundary regions that are adjacent on the spherical surface but not adjacent on the panoramic projection unfolded image in FIG. 6 are also processed according to the device described in this embodiment.
  • An embodiment of the present invention provides a panoramic image filtering method.
  • the common sub-region filtered pixels are determined by filtering the sub-regions corresponding to the sub-region corresponding to the plane boundary region that is encoded / decoded in the encoding / decoding sequence on the panoramic expansion projection image in the two plane boundary regions.
  • the pixel block b belongs to both the surface boundary region 8 and the surface boundary region 5, and the pixel block b is a common sub-region in the panoramic image filtering method.
  • the surface boundary region 8 is adjacent to the spherical surface but is not adjacent to the panoramic projection expanded image, and is 8 ′;
  • the surface boundary region 5 is adjacent to the spherical surface but is not adjacent to the panoramic projection expanded image. 5 '.
  • the pixel block b in the surface boundary region 5 is rotated clockwise by 90 ° and is placed above the pixel block b 2 ′ in the surface boundary region 5 ′. As shown in Figure 19.
  • the direction of the first-dimensional and second-dimensional filtering processes of the surface boundary region is determined according to the direction of the surface boundary region on the panoramic projection expansion image after the surface boundary region is centered.
  • the first-dimensional filtering processing direction of the non-plane boundary region is set to the vertical direction.
  • the pixel blocks b and b 1 ′ are first verticalized. Deblocking filtering in the vertical direction. Deblocking filtering in the vertical direction is performed on the pixel blocks b and b 2 ′. Then, the pixel blocks b in the area boundary area 8 and the pixel blocks b in the area boundary area 5 are horizontally processed. Filtering.
  • the filtered result of the pixel block b on the panoramic projection expanded image is determined by the pixel block b in the surface boundary region 5. This is because the area boundary area 5 is processed later than the area boundary area 8 in the encoding and decoding order.
  • the pixel blocks are sequentially deblocked and filtered in sequence, the pixel blocks below and to the right on the panoramic image will be processed later, that is, the surface boundary region 5 in FIG. 19 is later than the surface boundary region 8 Block filter processing. If the pixel blocks are deblocked in parallel, the filtered pixels in the common subregion are still determined by the boundary area of the surface processed later in the serial processing. This ensures serial processing and parallel processing. Consistency.
  • the present invention determines the direction and processing order of the first-dimensional and second-dimensional filtering processing of the pair of surface boundary regions, so that the encoding and decoding ends can match when performing filtering processing on the pair of surface boundary regions.
  • An embodiment of the present invention provides a method for filtering a panoramic image pixel block.
  • the position of the nominal point in the pixel block to be filtered along the direction of the spherical segmentation line is stretched to determine the The position of the nominal point corresponding to the corresponding nominal point on the other side of the spherical dividing line; determining the corresponding pixel block of the pixel block to be filtered on the other side of the spherical dividing line by the corresponding nominal point; Filtering the pixel block to be filtered and its corresponding pixel block to obtain a filtered result of the pixel block to be filtered.
  • the image in which the pixel block to be filtered is located in this embodiment is a 3x2 hexahedron format, as shown in FIG. 20.
  • Fig. 21 shows a pixel block B to be filtered and its corresponding pixel block B 'on the other side of the spherical segmentation line of the panoramic image. Assuming that the images have the same sampling rate in the X-Y coordinate system, it can be obtained from the format information that the images have different sampling rates in the X'-Y 'coordinate system in FIG. 21 and have the following relationship:
  • the nominal point is the geometric midpoint b of a row of pixels in the pixel block B to be filtered closest to the boundary of the surface.
  • the horizontal coordinate of the point is scaled to obtain the corresponding nominal point b in the pixel block B '. 'The ordinate.
  • the scaling method is as follows:
  • f -1 (x ′) represents the inverse function of f (x)
  • y in the formula for y ′ is ⁇ 1.
  • Figure 23 shows the corresponding nominal point b 'in the corresponding pixel block B'.
  • the coding information required for filtering such as the quantization step size, prediction mode (intra prediction or inter prediction) and other information comes from the nominal point b '.
  • the vertical coordinate of b ' is limited to obtain the whole pixel accuracy. Pixels that do not correspond to the nominal point are based on the position of the nominal point b' to obtain pixels at the interval of the whole pixel.
  • the coding information and pixel characteristics of the corresponding pixel block determine the filtering method.
  • the pixel block to be filtered and its corresponding pixel block are filtered to obtain the filtered pixels of the pixel block to be filtered.
  • An embodiment of the present invention provides a method for filtering a panoramic image pixel block.
  • the position of the nominal point in the pixel block to be filtered along the direction of the spherical segmentation line is stretched to determine the The position of the nominal point corresponding to the corresponding nominal point on the other side of the spherical dividing line; determining the corresponding pixel block of the pixel block to be filtered on the other side of the spherical dividing line by the corresponding nominal point; Filtering the pixel block to be filtered and its corresponding pixel block to obtain a filtered result of the pixel block to be filtered.
  • the image in which the pixel block to be filtered is located in this embodiment is a 3x2 hexahedron format, as shown in FIG. 20.
  • Fig. 21 shows a pixel block B to be filtered and its corresponding pixel block B 'on the other side of the spherical segmentation line of the panoramic image. Assuming that the images have the same sampling rate in the X-Y coordinate system, it can be obtained from the format information that the images have different sampling rates in the X'-Y 'coordinate system in FIG. 21 and have the following relationship:
  • the nominal point is the geometric midpoint b of a row of pixels in the pixel block B to be filtered closest to the boundary of the surface.
  • the horizontal coordinate of the point is scaled to obtain the corresponding nominal point b in the pixel block B ' 'The ordinate.
  • the scaling method is as follows:
  • f -1 (x ′) represents the inverse function of f (x).
  • Figure 24 shows the corresponding nominal point b 'in the corresponding pixel block B'.
  • the encoding information such as the quantization step size, prediction mode (intra prediction or inter prediction), etc. required for filtering comes from the nominal point b '.
  • Encoding unit which takes the limited precision of the ordinate of the corresponding nominal point b 'in B' to obtain the sub-pixel accuracy, such as 1/2 pixel accuracy, 1/4 pixel accuracy, 1/8 pixel accuracy, 1/16 pixel accuracy Wait.
  • the remaining pixels at the non-corresponding nominal points are obtained at the interval of whole pixels with reference to b '.
  • the pixels at the non-integrated pixels are obtained by interpolation of the surrounding whole pixel positions.
  • Information and pixel characteristics determine the filtering method.
  • the filtering in this embodiment refers specifically to deblocking filtering, and the determination filtering method refers to determining a filtering coefficient when deblocking filtering is performed.
  • the pixel block to be filtered and its corresponding pixel block are filtered according to the determined filter coefficients to obtain a filtered pixel of the pixel block to be filtered.
  • An embodiment of the present invention provides a method for filtering a panoramic image pixel block.
  • the position of the nominal point in the pixel block to be filtered along the direction of the spherical segmentation line is stretched to determine the The position of the nominal point corresponding to the corresponding nominal point on the other side of the spherical dividing line; determining the corresponding pixel block of the pixel block to be filtered on the other side of the spherical dividing line by the corresponding nominal point; Filtering the pixel block to be filtered and its corresponding pixel block to obtain a filtered result of the pixel block to be filtered.
  • the image in which the pixel block to be filtered is located in this embodiment is a 3x2 hexahedron format, as shown in FIG. 20.
  • Fig. 21 shows a pixel block B to be filtered and its corresponding pixel block B 'on the other side of the spherical segmentation line of the panoramic image. Assuming that the images have the same sampling rate in the X-Y coordinate system, it can be obtained from the format information that the images have different sampling rates in the X'-Y 'coordinate system in FIG. 21 and have the following relationship:
  • the nominal point is the pixel point b on the left side of the row of pixels in the pixel block B to be filtered closest to the surface boundary.
  • the horizontal coordinate of the point is scaled to obtain the corresponding nominal point b in the pixel block B '. 'The ordinate.
  • the scaling method is as follows:
  • f -1 (x ′) represents the inverse function of f (x).
  • Figure 26 shows the corresponding nominal point b 'in the corresponding pixel block B'. Since the coding unit C contains the largest number of pixels in B ', the coding information required for filtering such as quantization step size, prediction mode (intra-frame) (Prediction or inter prediction) and other information to encode unit C, and at the same time, the ordinate of the nominal point b 'is limited to obtain the whole pixel accuracy. Pixels that do not correspond to the nominal point are based on the position of the nominal point b' as the whole pixel.
  • An embodiment of the present invention provides a method for filtering a panoramic image pixel block.
  • the position of the nominal point in the pixel block to be filtered along the direction of the spherical segmentation line is stretched to determine the The position of the nominal point corresponding to the corresponding nominal point on the other side of the spherical dividing line; determining the corresponding pixel block of the pixel block to be filtered on the other side of the spherical dividing line by the corresponding nominal point; Filtering the pixel block to be filtered and its corresponding pixel block to obtain a filtered result of the pixel block to be filtered.
  • the image in which the pixel block to be filtered is located in this embodiment is a 3x2 hexahedron format, as shown in FIG. 20.
  • Fig. 21 shows a pixel block B to be filtered and its corresponding pixel block B 'on the other side of the spherical segmentation line of the panoramic image. Assuming that the images have the same sampling rate in the X-Y coordinate system, it can be obtained from the format information that the images have different sampling rates in the X'-Y 'coordinate system in FIG. 21 and have the following relationship:
  • the nominal point is the pixel point b on the left side of the row of pixels in the pixel block B to be filtered closest to the surface boundary.
  • the horizontal coordinate of the point is scaled to obtain the corresponding nominal point b in the pixel block B '. 'The ordinate.
  • the scaling method is as follows:
  • f -1 (x ′) represents the inverse function of f (x).
  • Figure 26 shows the corresponding nominal point b 'in the corresponding pixel B' on the panoramic image of the pixel block B to be filtered.
  • the coding unit C contains the largest number of pixels in B ', the coding information required for filtering such as quantization Step length, prediction mode (intra prediction or inter prediction) and other information to encode unit C, and at the same time, the vertical coordinate of the nominal point b 'is limited to obtain the sub-pixel accuracy, such as 1/2 pixel accuracy, 1/4 pixel Accuracy, 1/8 pixel accuracy, 1/16 pixel accuracy, etc.
  • Pixels at non-corresponding nominal points are obtained at the full pixel interval with reference to the position of the nominal point b ', and the filtering method is determined based on the coding information and pixel characteristics of the pixel block to be filtered and its corresponding pixel block.
  • the corresponding pixel block is filtered to obtain the filtered pixel block and the corresponding filtered pixel block.
  • An embodiment of the present invention provides a method for filtering a panoramic image pixel block.
  • the position of the nominal point in the pixel block to be filtered along the direction of the spherical segmentation line is stretched to determine the The position of the nominal point corresponding to the corresponding nominal point on the other side of the spherical dividing line; determining the corresponding pixel block of the pixel block to be filtered on the other side of the spherical dividing line by the corresponding nominal point; Filtering the pixel block to be filtered and its corresponding pixel block to obtain a filtered result of the pixel block to be filtered.
  • the image in which the pixel block to be filtered is located in this embodiment is a 3x2 hexahedron format, as shown in FIG. 20.
  • Fig. 21 shows a pixel block B to be filtered and its corresponding pixel block B 'on the other side of the spherical segmentation line of the panoramic image. Assuming that the images have the same sampling rate in the X-Y coordinate system, it can be obtained from the format information that the images have different sampling rates in the X'-Y 'coordinate system in FIG. 21 and have the following relationship:
  • the nominal point is the geometric midpoint b of a row of pixels in the pixel block B to be closest to the boundary of the surface, and the vertical coordinate corresponding to the nominal point b 'in the pixel block B' is calculated based on the horizontal coordinate of the point. .
  • the calculation method is as follows:
  • f -1 (x ′) represents the inverse function of f (x).
  • FIG. 23 shows the corresponding nominal point b ′ in the corresponding pixel block B ′.
  • the ordinate of the corresponding nominal point b ′ in B ′ is limited to obtain the sub-pixel accuracy, such as 1/2 pixel accuracy, 1 / 4 pixel accuracy, 1/8 pixel accuracy, 1/16 pixel accuracy, etc.
  • the pixels at the other non-corresponding nominal point locations are obtained at the whole pixel interval with reference to b '.
  • the pixels at the non-integer pixel location are obtained by interpolation of the surrounding whole pixel locations.
  • the to-be-filtered pixel block and its corresponding pixel block are filtered to obtain filtered pixels of the to-be-filtered pixel block.
  • An embodiment of the present invention provides a method for filtering a panoramic image pixel block.
  • the position of the nominal point in the pixel block to be filtered along the direction of the spherical segmentation line is stretched to determine the The position of the nominal point corresponding to the corresponding nominal point on the other side of the spherical dividing line; determining the corresponding pixel block of the pixel block to be filtered on the other side of the spherical dividing line by the corresponding nominal point; Filtering the pixel block to be filtered and its corresponding pixel block to obtain a filtered result of the pixel block to be filtered.
  • the image in which the pixel block to be filtered is located in this embodiment is a 3x2 hexahedron format, as shown in FIG. 20.
  • Fig. 21 shows a pixel block B to be filtered and its corresponding pixel block B 'on the other side of the spherical segmentation line of the panoramic image. Assuming that the images have the same sampling rate in the X-Y coordinate system, it can be obtained from the format information that the images have different sampling rates in the X'-Y 'coordinate system in FIG. 21 and have the following relationship:
  • the nominal point is the geometric midpoint b of a row of pixels in the pixel block B to be closest to the boundary of the surface, and the vertical coordinate corresponding to the nominal point b 'in the pixel block B' is calculated based on the horizontal coordinate of the point. .
  • the calculation method is as follows:
  • f -1 (x ′) represents the inverse function of f (x).
  • FIG. 23 shows the corresponding nominal point b ′ in the corresponding pixel block B ′.
  • the ordinate of the corresponding nominal point b ′ in B ′ is limited to obtain the sub-pixel accuracy, such as 1/2 pixel accuracy, 1 / 4 pixel accuracy, 1/8 pixel accuracy, 1/16 pixel accuracy, etc.
  • the pixels at the other non-corresponding nominal point locations are obtained at the whole pixel interval with reference to b '.
  • the pixels at the non-integer pixel location are obtained by interpolation of the surrounding whole pixel locations.
  • the to-be-filtered pixel block and its corresponding pixel block are filtered to obtain filtered pixels of the to-be-filtered pixel block.
  • An embodiment of the present invention provides a method for filtering a panoramic image pixel block.
  • the position of the nominal point in the pixel block to be filtered along the direction of the spherical segmentation line is stretched to determine The position of the nominal point corresponding to the corresponding nominal point on the other side of the spherical dividing line; determining the corresponding pixel block of the pixel block to be filtered on the other side of the spherical dividing line by the corresponding nominal point; Filtering the pixel block to be filtered and its corresponding pixel block to obtain a filtered result of the pixel block to be filtered.
  • the image in which the pixel block to be filtered is located in this embodiment is a 3x2 hexahedron format, as shown in FIG. 20.
  • Fig. 21 shows a pixel block B to be filtered and its corresponding pixel block B 'on the other side of the spherical segmentation line of the panoramic image. Assuming that the images have the same sampling rate in the X-Y coordinate system, it can be obtained from the format information that the images have different sampling rates in the X'-Y 'coordinate system in FIG. 21 and have the following relationship:
  • the nominal point is the geometric midpoint b of a row of pixels in the pixel block B to be closest to the boundary of the surface, and the vertical coordinate corresponding to the nominal point b 'in the pixel block B' is calculated based on the horizontal coordinate of the point. .
  • the calculation method is as follows:
  • f -1 (x ′) represents the inverse function of f (x).
  • FIG. 23 shows the corresponding nominal point b 'in the corresponding pixel block B', and the ordinate of the corresponding nominal point b 'in B' is limited to obtain the precision of the whole pixel.
  • the pixels at the other non-corresponding nominal points are obtained at the interval of whole pixels with reference to b '.
  • the to-be-filtered pixel block and its corresponding pixel block are filtered to obtain filtered pixels of the to-be-filtered pixel block.
  • An embodiment of the present invention provides a method for filtering a panoramic image pixel block.
  • the position of the nominal point in the pixel block to be filtered along the direction of the spherical segmentation line is stretched to determine the The position of the nominal point corresponding to the corresponding nominal point on the other side of the spherical dividing line; determining the corresponding pixel block of the pixel block to be filtered on the other side of the spherical dividing line by the corresponding nominal point; Filtering the pixel block to be filtered and its corresponding pixel block to obtain a filtered result of the pixel block to be filtered.
  • the image in which the pixel block to be filtered is located in this embodiment is a 3x2 hexahedron format, as shown in FIG. 20.
  • Fig. 21 shows a pixel block B to be filtered and its corresponding pixel block B 'on the other side of the spherical segmentation line of the panoramic image. Assuming that the images have the same sampling rate in the X-Y coordinate system, it can be obtained from the format information that the images have different sampling rates in the X'-Y 'coordinate system in FIG. 21 and have the following relationship:
  • the nominal point is the geometric midpoint b of a row of pixels in the pixel block B to be closest to the boundary of the surface, and the vertical coordinate corresponding to the nominal point b 'in the pixel block B' is calculated based on the horizontal coordinate of the point. .
  • the calculation method is as follows:
  • f -1 (x ′) represents the inverse function of f (x).
  • FIG. 23 shows the corresponding nominal point b ′ in the corresponding pixel block B ′.
  • the ordinate of the corresponding nominal point b ′ in B ′ is limited to obtain the sub-pixel accuracy, such as 1/2 pixel accuracy, 1 / 4 pixel accuracy, 1/8 pixel accuracy, 1/16 pixel accuracy, etc.
  • the pixels at the other non-corresponding nominal point locations are obtained at the whole pixel interval with reference to b '.
  • the pixels at the non-integer pixel location are obtained by interpolation of the surrounding whole pixel locations.
  • the to-be-filtered pixel block and its corresponding pixel block are filtered to obtain filtered pixels of the to-be-filtered pixel block.
  • An embodiment of the present invention provides a method for filtering a panoramic image pixel block.
  • the corresponding pixel of the pixel block to be filtered on the other side of the spherical segmentation line is determined.
  • the pixel block to be filtered and its corresponding pixel block are filtered to obtain the filtered result of the pixel block to be filtered.
  • Fig. 27 shows a pixel block B to be filtered and its corresponding pixel block B 'on the panoramic image.
  • the filtering method is determined according to the coding information of the pixel block to be filtered and its corresponding pixel block, wherein the motion information in the coding information of the pixel block to be filtered and its corresponding pixel block does not need to be obtained, and the pixel block to be filtered and its corresponding pixel block are considered The motion information is always equal. Filtering the pixel block to be filtered and its corresponding pixel block to obtain filtered pixels of the pixel block to be filtered.
  • An embodiment of the present invention provides a method for filtering a panoramic image pixel block.
  • the corresponding pixel of the pixel block to be filtered on the other side of the spherical segmentation line is determined.
  • the pixel block to be filtered and its corresponding pixel block are filtered to obtain the filtered result of the pixel block to be filtered.
  • Fig. 28 shows a pixel block B to be filtered and its corresponding pixel block B 'on a latitude and longitude map format.
  • a filtering method is determined according to the coding information of the pixel block to be filtered and its corresponding pixel block, and it is considered that the motion information in the coding information of the pixel block to be filtered and its corresponding pixel block is always equal. Filtering the pixel block to be filtered and its corresponding pixel block to obtain filtered pixels of the pixel block to be filtered.
  • An embodiment of the present invention provides a method for filtering a panoramic image pixel block.
  • the position of the nominal point in the pixel block to be filtered along the direction of the spherical segmentation line is stretched to determine the The position of the nominal point corresponding to the corresponding nominal point on the other side of the spherical dividing line; determining the corresponding pixel block of the pixel block to be filtered on the other side of the spherical dividing line by the corresponding nominal point; Filtering the pixel block to be filtered and its corresponding pixel block to obtain a filtered result of the pixel block to be filtered.
  • the filtering in this embodiment refers to deblocking filtering. Before filtering the pixel block to be filtered and its corresponding pixel block, it is necessary to determine the corresponding filtering method. When determining the filtering method, set the boundary strength of the pixel block to be filtered and its corresponding pixel block to 1, and determine the filtering method without obtaining The coding information of the filtered pixel block and its corresponding pixel block. Instead, the quantization step size of the image where the pixel block to be filtered is used is the quantization step size of the pixel block to be filtered, and the quantization step size of the image where the corresponding pixel block is located is used as the quantization of the corresponding pixel block. Step size.
  • the coding method is determined according to the boundary strength, the quantization step size, and the pixel characteristics in the pixel block to be filtered and its corresponding pixel block.
  • the pixel block to be filtered and its corresponding pixel block are filtered to obtain the filtered pixels of the pixel block to be filtered.
  • An embodiment of the present invention provides a method for filtering a panoramic image pixel block.
  • the position of the nominal point in the pixel block to be filtered along the direction of the spherical segmentation line is stretched to determine the The position of the nominal point corresponding to the corresponding nominal point on the other side of the spherical dividing line; determining the corresponding pixel block of the pixel block to be filtered on the other side of the spherical dividing line by the corresponding nominal point; Filtering the pixel block to be filtered and its corresponding pixel block to obtain a filtered result of the pixel block to be filtered.
  • the filtering method Before filtering the pixel block to be filtered and its corresponding pixel block, it is necessary to determine the corresponding filtering method. When determining the filtering method, set the boundary strength of the pixel block to be filtered and its corresponding pixel block to 2, and determine the filtering method without obtaining Filter coding information of pixel blocks and their corresponding pixel blocks. Use the quantization step size of the image where the pixel block to be filtered is used as the quantization step size of the pixel block to be filtered, and the quantization step size of the image where the corresponding pixel block is located as the quantization step size of the corresponding pixel block.
  • the encoding method is determined according to the boundary strength, the quantization step size, and the pixel characteristics in the pixel block to be filtered and its corresponding pixel block.
  • the pixel block to be filtered and its corresponding pixel block are filtered to obtain the filtered pixels of the pixel block to be filtered.
  • An embodiment of the present invention provides a method for filtering a panoramic image pixel block.
  • the position of the nominal point in the pixel block to be filtered along the direction of the spherical segmentation line is stretched to determine the The position of the nominal point corresponding to the corresponding nominal point on the other side of the spherical dividing line; determining the corresponding pixel block of the pixel block to be filtered on the other side of the spherical dividing line by the corresponding nominal point; Filtering the pixel block to be filtered and its corresponding pixel block to obtain a filtered result of the pixel block to be filtered.
  • the boundary strength of the pixel block to be filtered and its corresponding pixel block is determined according to the coding information of the two pixel blocks. If the prediction mode in the coding information of at least one block in the pixel block is intra prediction, the boundary strength is 2; otherwise, the boundary strength is 1.
  • the quantization step size of the image where the pixel block to be filtered is used is used as the quantization step size of the pixel block to be filtered, and the quantization step size of the image where the corresponding pixel block is located is used as the quantization step size of the corresponding pixel block.
  • the encoding method is determined according to the boundary strength, the quantization step size, and the pixel characteristics in the pixel block to be filtered and its corresponding pixel block.
  • the pixel block to be filtered and its corresponding pixel block are filtered to obtain the filtered pixels of the pixel block to be filtered.
  • An embodiment of the present invention provides a method for filtering a panoramic image pixel block.
  • the position of the nominal point in the pixel block to be filtered along the direction of the spherical segmentation line is stretched to determine the The position of the nominal point corresponding to the corresponding nominal point on the other side of the spherical dividing line; determining the corresponding pixel block of the pixel block to be filtered on the other side of the spherical dividing line by the corresponding nominal point; Filtering the pixel block to be filtered and its corresponding pixel block to obtain a filtered result of the pixel block to be filtered.
  • the boundary strength of the pixel block to be filtered and its corresponding pixel block is determined according to the coding information of the two pixel blocks. If the prediction mode in the coding information of at least one block in the pixel block is intra prediction, the boundary strength is 2; otherwise, the boundary strength is 1.
  • the quantization step required to determine the filtered pixels is taken from the coding information of the pixel block to be filtered and its corresponding pixel block.
  • the encoding method is determined according to the boundary strength, the quantization step size, and the pixel characteristics in the pixel block to be filtered and its corresponding pixel block.
  • the pixel block to be filtered and its corresponding pixel block are filtered to obtain the filtered pixels of the pixel block to be filtered.
  • An embodiment of the present invention provides a method for filtering a panoramic image pixel block.
  • the position of the nominal point in the pixel block to be filtered along the direction of the spherical segmentation line is stretched to determine the The position of the nominal point corresponding to the corresponding nominal point on the other side of the spherical dividing line; determining the corresponding pixel block of the pixel block to be filtered on the other side of the spherical dividing line by the corresponding nominal point; Filtering the pixel block to be filtered and its corresponding pixel block to obtain a filtered result of the pixel block to be filtered.
  • the boundary strength of the pixel block to be filtered and its corresponding pixel block is determined according to the coding information of the two pixel blocks. If the prediction mode in the coding information of at least one block in the pixel block is intra prediction, the boundary strength is 2; otherwise, the boundary strength is 1.
  • the quantization step size required to determine the filtering method is not directly taken from the pixel block to be filtered and its corresponding pixel block, but the quantization step size of the coding region where the pixel block to be filtered is located and the larger quantization step size of the pixel block to be filtered are used.
  • the value is used as the quantization step size of the pixel block to be filtered; the quantization step size of the coding region where the corresponding pixel block is located and the quantization step size of the corresponding pixel block are used as the quantization step size of the corresponding pixel block.
  • the encoding method is determined according to the boundary strength, the quantization step size, and the pixel characteristics in the pixel block to be filtered and its corresponding pixel block.
  • the pixel block to be filtered and its corresponding pixel block are filtered to obtain the filtered pixels of the pixel block to be filtered.
  • An embodiment of the present invention provides a device for filtering a panoramic image pixel block.
  • Corresponding nominal point position derivation module The input of this module is the position information of the nominal point in the pixel block to be filtered immediately to the side of the spherical segmentation line of the panoramic image and the projection format information corresponding to the image where the pixel block to be filtered is located; the output is The position of the nominal point in the panoramic image corresponding to the nominal point; according to the projection format information corresponding to the image of the pixel block to be filtered on the side of the spherical segmentation line of the panoramic image, the nominal point in the pixel block to be filtered is Performing a telescoping transformation on a position in the direction of the spherical dividing line to determine a position of the nominal point corresponding to a corresponding nominal point on the other side of the spherical dividing line;
  • FIG. 29 is a schematic diagram of a panoramic image pixel block filtering device according to this embodiment.
  • the position of the corresponding nominal point is taken with limited precision to obtain the whole pixel position.
  • Corresponding pixel block position derivation module The module input is the position of the corresponding nominal point, and the output is the position of the corresponding pixel block in the panoramic image of the pixel block to be filtered; this module is used to determine the corresponding pixel block according to the position of the corresponding nominal point. The location of the remaining non-corresponding nominal points;
  • the method for determining the position of the non-nominal point in this module is: determining the position of the non-corresponding nominal point at the whole pixel interval based on the corresponding nominal point in the corresponding pixel block.
  • Pixel filtering module The input of this module is the position of the pixel block to be filtered and the position of the corresponding pixel block, and the output is the filtered pixel of the pixel block to be filtered. This module is used to filter the pixel block to be filtered and its corresponding pixel block. The filtered result of the pixel block to be filtered is obtained.
  • This module needs to obtain the coding information of the pixel block to be filtered and its corresponding pixel block, determine the filtering method according to the coding information of the pixel block to be filtered and its corresponding pixel block, and the pixels, and filter the pixel block to be filtered and its corresponding pixel block to obtain the filter to be filtered Pixel block filtered results.
  • An embodiment of the present invention provides a device for filtering a panoramic image pixel block.
  • Corresponding pixel block position derivation module The module inputs the position information of the pixel block to be filtered immediately adjacent to the spherical segmentation line side of the panoramic image and the projection format information corresponding to the image where the pixel block to be filtered is located, and outputs the corresponding pixel block to be filtered.
  • a filtering method determining module is used to determine a filtering method.
  • the input is the position information of the pixel block to be filtered and its corresponding pixel block.
  • the filtering method is determined according to the coding information of the pixel block to be filtered and its corresponding pixel block.
  • the filtering method is determined as follows: One type: the motion information in the coding information of the pixel block to be filtered and its corresponding pixel block is considered to be the same, or the motion information in the coding information of the pixel block to be filtered and its corresponding pixel block is not used, and the output is The filtering method determined by the module;
  • a pixel filtering module is configured to filter the pixel block to be filtered and its corresponding pixel block to obtain a filtered result of the pixel block to be filtered.
  • FIG. 30 is a schematic diagram of a panoramic image pixel block filtering device according to this embodiment.
  • the position is passed to the filtering method determination module.
  • the filtering method determination module determines the filtering method and passes the determined filtering method to the pixel block to be filtered.
  • the pixel filtering module is based on The input filtering method filters the input pixel block to obtain the filtered result of the pixel block to be filtered.
  • An embodiment of the present invention provides a device for filtering a panoramic image pixel block.
  • Corresponding nominal point position derivation module The input of this module is the position information of the nominal point in the pixel block to be filtered immediately to the side of the spherical segmentation line of the panoramic image and the projection format information corresponding to the image where the pixel block to be filtered is located; the output is The position of the nominal point in the panoramic image corresponding to the nominal point; according to the projection format information corresponding to the image of the pixel block to be filtered on the side of the spherical segmentation line of the panoramic image, the nominal point in the pixel block to be filtered is Performing a telescoping transformation on a position in the direction of the spherical dividing line to determine a position of the nominal point corresponding to a corresponding nominal point on the other side of the spherical dividing line;
  • Corresponding pixel block position derivation module The module input is the position of the corresponding nominal point, and the output is the position of the corresponding pixel block in the panoramic image of the pixel block to be filtered; this module is used to determine the corresponding pixel block according to the position of the corresponding nominal point. The location of the remaining non-corresponding nominal points;
  • Pixel filtering module The input of this module is the position of the pixel block to be filtered and the position of the corresponding pixel block, and the output is the filtered pixel of the pixel block to be filtered. This module is used to filter the pixel block to be filtered and its corresponding pixel block. The filtered result of the pixel block to be filtered is obtained.
  • the corresponding nominal point position derivation module includes a corresponding nominal point position rounding module, which takes the limited precision of the corresponding nominal point coordinates to obtain the sub-pixel accuracy, such as 1/2 pixel accuracy, 1 / 4 pixel accuracy, 1/8 pixel accuracy, 1/16 pixel accuracy, etc.
  • the corresponding pixel block position derivation module also includes a non-corresponding nominal point position derivation module, as shown in FIG. 32.
  • the non-corresponding nominal point position derivation module determines the position of the non-corresponding nominal point by taking the corresponding nominal point in the corresponding pixel block as a reference and determining the position of the non-corresponding nominal point at an entire pixel interval.
  • the pixel filtering module further includes an encoding information acquisition module and a filtering method determination module, as shown in FIG. 33.
  • the encoding information acquisition module obtains the encoding information of the pixel block to be filtered and its corresponding pixel block according to the input position of the pixel block to be filtered and its corresponding pixel block; the filtering method determination module is based on the encoding of the pixel block to be filtered and its corresponding pixel block.
  • Information and pixel characteristics determine the filtering method. Filter the pixel block to be filtered and its corresponding block according to the determined filtering method, and obtain the filtered result of the pixel to be filtered.
  • An embodiment of the present invention provides a device for filtering a panoramic image pixel block.
  • Corresponding nominal point position derivation module The input of this module is the position information of the nominal point in the pixel block to be filtered immediately to the side of the spherical segmentation line of the panoramic image and the projection format information corresponding to the image where the pixel block to be filtered is located; the output is The position of the nominal point in the panoramic image corresponding to the nominal point; according to the projection format information corresponding to the image of the pixel block to be filtered on the side of the spherical segmentation line of the panoramic image, the nominal point in the pixel block to be filtered is Performing a telescoping transformation on a position in the direction of the spherical dividing line to determine a position of the nominal point corresponding to a corresponding nominal point on the other side of the spherical dividing line;
  • Corresponding pixel block position derivation module The module input is the position of the corresponding nominal point, and the output is the position of the corresponding pixel block in the panoramic image of the pixel block to be filtered; this module is used to determine the corresponding pixel block according to the position of the corresponding nominal point. The location of the remaining non-corresponding nominal points;
  • Pixel filtering module The input of this module is the position of the pixel block to be filtered and the position of the corresponding pixel block, and the output is the filtered pixel of the pixel block to be filtered. This module is used to filter the pixel block to be filtered and its corresponding pixel block. The filtered result of the pixel block to be filtered is obtained.
  • the filtering module before filtering the pixel block to be filtered and its corresponding pixel block, it is necessary to determine the corresponding filtering method.
  • determine the filtering method set the boundary strength of the pixel block to be filtered and its corresponding pixel block to 2 to determine the filtering. The method does not need to obtain the coding information of the pixel block to be filtered and its corresponding pixel block.
  • the quantization step size of the image where the pixel block to be filtered is used is used as the quantization step size of the pixel block to be filtered, and the quantization step size of the image where the corresponding pixel block is located is used as the quantization step size of the corresponding pixel block.
  • the encoding method is determined according to the boundary strength, the quantization step size, and the pixel characteristics in the pixel block to be filtered and its corresponding pixel block.
  • the pixel block to be filtered and its corresponding pixel block are filtered to obtain the filtered pixels of the pixel block to be filtered.
  • An embodiment of the present invention provides a device for filtering a panoramic image pixel block.
  • Corresponding nominal point position derivation module The input of this module is the position information of the nominal point in the pixel block to be filtered immediately to the side of the spherical segmentation line of the panoramic image and the projection format information corresponding to the image where the pixel block to be filtered is located; the output is The position of the nominal point in the panoramic image corresponding to the nominal point; according to the projection format information corresponding to the image of the pixel block to be filtered on the side of the spherical segmentation line of the panoramic image, the nominal point in the pixel block to be filtered is Performing a telescoping transformation on a position in the direction of the spherical dividing line to determine a position of the nominal point corresponding to a corresponding nominal point on the other side of the spherical dividing line;
  • Corresponding pixel block position derivation module The module input is the position of the corresponding nominal point, and the output is the position of the corresponding pixel block in the panoramic image of the pixel block to be filtered; this module is used to determine the corresponding pixel block according to the position of the corresponding nominal point The location of the remaining non-corresponding nominal points;
  • Pixel filtering module The input of this module is the position of the pixel block to be filtered and the position of the corresponding pixel block, and the output is the filtered pixel of the pixel block to be filtered. This module is used to filter the pixel block to be filtered and its corresponding pixel block. The filtered result of the pixel block to be filtered is obtained.
  • the filtering module before filtering the pixel block to be filtered and its corresponding pixel block, it is necessary to determine the corresponding filtering method.
  • determine the filtering method set the boundary strength of the pixel block to be filtered and its corresponding pixel block to 1 to determine the filtering. The method does not need to obtain the coding information of the pixel block to be filtered and its corresponding pixel block.
  • the quantization step size of the image where the pixel block to be filtered is used is used as the quantization step size of the pixel block to be filtered, and the quantization step size of the image where the corresponding pixel block is located is used as the quantization step size of the corresponding pixel block.
  • the encoding method is determined according to the boundary strength, the quantization step size, and the pixel characteristics in the pixel block to be filtered and its corresponding pixel block.
  • the pixel block to be filtered and its corresponding pixel block are filtered to obtain the filtered pixels of the pixel block to be filtered.
  • An embodiment of the present invention provides a device for filtering a panoramic image pixel block.
  • Corresponding pixel block position derivation module The module inputs the position information of the pixel block to be filtered immediately adjacent to the spherical segmentation line side of the panoramic image and the projection format information corresponding to the image where the pixel block to be filtered is located, and outputs the corresponding pixel block to be filtered.
  • a coding information acquisition module configured to obtain coding information of the pixel block to be filtered and its corresponding pixel block, wherein the motion information in the coding information of the pixel block to be filtered and its corresponding pixel block is always considered to be the same;
  • a pixel filtering module is configured to filter the pixel block to be filtered and its corresponding pixel block to obtain a filtered result of the pixel block to be filtered.
  • the motion information in the coding information of the pixel block to be filtered and its corresponding pixel block need not be acquired, and it is considered that the motion information of the pixel block to be filtered and its corresponding pixel block is always equal.
  • An embodiment of the present invention provides a device for filtering a panoramic image pixel block.
  • Corresponding nominal point position derivation module The input of this module is the position information of the nominal point in the pixel block to be filtered immediately to the side of the spherical segmentation line of the panoramic image and the projection format information corresponding to the image where the pixel block to be filtered is located; the output is The position of the nominal point in the panoramic image corresponding to the nominal point; according to the projection format information corresponding to the image of the pixel block to be filtered on the side of the spherical segmentation line of the panoramic image, the nominal point in the pixel block to be filtered is Performing a telescoping transformation on a position in the direction of the spherical dividing line to determine a position of the nominal point corresponding to a corresponding nominal point on the other side of the spherical dividing line;
  • Corresponding pixel block position derivation module The module input is the position of the corresponding nominal point, and the output is the position of the corresponding pixel block in the panoramic image of the pixel block to be filtered; this module is used to determine the corresponding pixel block according to the position of the corresponding nominal point. The location of the remaining non-corresponding nominal points;
  • Pixel filtering module The input of this module is the position of the pixel block to be filtered and the position of the corresponding pixel block, and the output is the filtered pixel of the pixel block to be filtered. This module is used to filter the pixel block to be filtered and its corresponding pixel block. The filtered result of the pixel block to be filtered is obtained.
  • the coordinates of the corresponding nominal point are rounded to obtain the whole pixel position; in the corresponding pixel block position derivation module, the positions of the remaining non-nominal points in the corresponding pixel block are corresponding to the nominal point. As a reference, it is acquired at the whole pixel interval.
  • the filtering module before filtering the pixel block to be filtered and its corresponding pixel block, it is necessary to determine the corresponding filtering method.
  • determine the filtering method set the boundary strength of the pixel block to be filtered and its corresponding pixel block to 1 to determine the filtering. The method does not need to obtain the coding information of the pixel block to be filtered and its corresponding pixel block.
  • the quantization step size of the image where the pixel block to be filtered is used is used as the quantization step size of the pixel block to be filtered, and the quantization step size of the image where the corresponding pixel block is located is used as the quantization step size of the corresponding pixel block.
  • the encoding method is determined according to the boundary strength, the quantization step size, and the pixel characteristics in the pixel block to be filtered and its corresponding pixel block.
  • the pixel block to be filtered and its corresponding pixel block are filtered to obtain the filtered pixels of the pixel block to be filtered.
  • FIG. 34 shows the artifacts in a viewing area of the unfiltered panoramic image.
  • the viewing area generated by filtering the pixel blocks of the panoramic image to be filtered by using the panoramic image pixel block filtering device of this embodiment is shown in FIG. 35 It can be seen that the artifacts caused by encoding are greatly reduced.
  • An embodiment of the present invention provides a method for filtering a panoramic image pixel block.
  • the pixel block to be filtered immediately to the side of the spherical segmentation line of the panoramic image determines the corresponding pixel of the pixel block to be filtered on the other side of the spherical segmentation line Block position; determine a filtering method according to the coding information of the pixel block to be filtered and its corresponding pixel block, and determine the filtering method as one of the following: consider the motion in the coding information of the pixel block to be filtered and its corresponding pixel block The same information, or not using the motion information in the coding information of the pixel block to be filtered and its corresponding pixel block; filtering the pixel block to be filtered and its corresponding pixel block to obtain the filtered result of the pixel block to be filtered .
  • the foregoing process of determining a filtering method according to the coding information of the pixel block to be filtered and its corresponding pixel block is described in more detail.
  • determining the filtering method it is necessary to determine the boundary strength of the pixel block to be filtered and its corresponding pixel block.
  • the boundary strength the motion information in the encoded information is not used or the motion information in the encoded information is considered equal. If the prediction mode in the coding information of at least one block is intra prediction, the boundary strength is 2; otherwise, the boundary strength is 1.
  • the QP in the encoded information is used to determine the decision threshold of the filtering method.
  • the coding method is determined according to the boundary strength, the decision threshold, and the pixel characteristics in the pixel block to be filtered and its corresponding pixel block.
  • the pixel block to be filtered and its corresponding pixel block are filtered to obtain the filtered pixels of the pixel block to be filtered.
  • An embodiment of the present invention provides a method for filtering a panoramic image pixel block.
  • the corresponding pixel of the pixel block to be filtered on the other side of the spherical segmentation line is determined.
  • Block position determine a filtering method according to the coding information of the pixel block to be filtered and its corresponding pixel block, and determine the filtering method as one of the following: consider the motion in the coding information of the pixel block to be filtered and its corresponding pixel block The same information, or not using the motion information in the coding information of the pixel block to be filtered and its corresponding pixel block; filtering the pixel block to be filtered and its corresponding pixel block to obtain the filtered result of the pixel block to be filtered .
  • the foregoing process of determining a filtering method according to the coding information of the pixel block to be filtered and its corresponding pixel block is described in more detail.
  • determining the filtering method it is necessary to determine the boundary strength of the pixel block to be filtered and its corresponding pixel block.
  • the boundary strength is set to be fixed.
  • the value 1 or a fixed value 2 is used, and the QP in the encoded information is used to determine the decision threshold of the filtering method.
  • the coding method is determined according to the boundary strength, the decision threshold, and the pixel characteristics in the pixel block to be filtered and its corresponding pixel block.
  • the pixel block to be filtered and its corresponding pixel block are filtered to obtain the filtered pixels of the pixel block to be filtered.
  • An embodiment of the present invention provides a method for filtering a panoramic image pixel block.
  • the corresponding pixel of the pixel block to be filtered on the other side of the spherical segmentation line is determined.
  • Block position determine a filtering method according to the coding information of the pixel block to be filtered and its corresponding pixel block, and determine the filtering method as one of the following: consider the motion in the coding information of the pixel block to be filtered and its corresponding pixel block The same information, or not using the motion information in the coding information of the pixel block to be filtered and its corresponding pixel block; filtering the pixel block to be filtered and its corresponding pixel block to obtain the filtered result of the pixel block to be filtered .
  • Figure 36 (b) shows another method of boundary strength judgment, that is, the boundary strength is fixed at 2 or fixed at 1.
  • the decision threshold of each filtering method needs to be determined according to the QP in the coding information of the pixel block to be filtered and its corresponding block. Determining different thresholds for different quantization situations can better improve the filtering effect. .
  • Figure 37 shows the boundary strength decision process in the deblocking filtering technology in HEVC.
  • the number of bits required in the judgment are: Intra / inter prediction flag, 1 bit; non-zero transform coefficient flag, 1 bit; motion information: motion vector (MVx, MVy) where MVx and MVy each require 16 bits, and the index of the reference frame: 4 bits, so there is a total of two in hypothetical prediction 72bit of motion information is required.
  • Intra / inter prediction flag 1 bit
  • motion information motion vector (MVx, MVy) where MVx and MVy each require 16 bits, and the index of the reference frame: 4 bits
  • the coding information is stored in blocks, for example, the coding information is stored in blocks of 4 * 4 size.
  • the present invention saves 73bit or 74bit for each 4 * 4 block.
  • the size occupied by the number of pixels in each 4 * 4 block is 16 * bitdepth, where bitdepth represents the number of bits used by each pixel, such as 8bit or 10bit.
  • bitdepth represents the number of bits used by each pixel, such as 8bit or 10bit.
  • 8bit the pixels of a 4 * 4 size pixel block are used
  • the total number of bits is 128 bits, plus a total of 202 bits of coding information.
  • the deblocking filtering technology in HEVC uses 202 bits for each 4 * 4 block, while the method in the embodiment of the present invention uses only 128 bits or 129 bits. Compared to the deblocking filtering technology, it saves 36% of the line cache.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Compression Or Coding Systems Of Tv Signals (AREA)
  • Image Processing (AREA)

Abstract

本发明公开了一种全景图像滤波方法及装置。根据全景投影展开图像的格式信息确定在全景投影展开图像中存在的在全景球面上相邻、但在投影展开图像上不相邻的面边界区域对的位置,按照所述面边界区域对中一个特定的面边界区域在全景投影展开图像上的方向,确定面边界区域对中第一维和第二维滤波处理的方向,对所述面边界区域对依次进行第一维和第二维滤波处理,所述面边界区域对确定第一维和第二维滤波处理方向的方法与非面边界区域在全景投影展开图像上确定第一维和第二维滤波处理的方向一致。本发明可以减少编码图像的编码伪像和面边界的不连续性,提高了编码图像的主观质量,同时保证了编解码端的匹配。

Description

一种全景图像滤波方法及装置 技术领域
本发明涉及全景图像处理领域,特别涉及全景图像像素块滤波方法及装置。
背景技术
全景图像指空间中一个观察点四周所有的场景,由这个观察点所能接收到的所有光线构成,球面可以描述观察点四周的所有场景。由于球面图像难以存储及编码,所以全景图像的球面通过定义特定的投影展开方法,将球面图像展开为平面图像,该平面图像称为全景投影展开图像,所述的投影展开方法一般由某一全景图像展开格式信息确定。
立方体是目前最常见的全景图像投影展开格式之一,如图1所示,利用中心投影的方法将球面投影到其对应的外接立方体的六个面上,立方体体的八个顶点分别为A、B、C、D、E、F、G、H,然后将立方体展开,其中左侧面、前面以及右侧面水平排布,而上面、后面以及下面垂直排布。由于左侧面、前面以及右侧面展示的是立方体的外表面,而图1中上面、后面以及下面的垂直排布展示的立方体的内表面,所以需要将上面、后面以及下面的垂直排布左右翻转以表示立方体的外表面,如图2所示,将上面、后面以及下面左右翻转后再顺时针旋转90°,将立方体的6个面紧凑排布成3x2的矩形图像,图2仅展示出立方体6个面的一种排布情况,立方体的6个面也可以不按照这种方式排布。
除了立方体格式以外,还有其他相对于立方体格式改进的格式,目前表达效率较高的全景图像投影展开格式有EAC、PAU以及HEC等。
将紧邻立方体6个面的边界的区域称为面边界区域,将球面投影展开为全景投影展开图像后,原先在球面上相邻的面边界区域在全景投影展开图像上可能不相邻,如图3所示,面边界区域1和1’在立方体上有共同的边界AE,而在全景投影展开图像上面边界区域1和1’不相邻,图3中数字a和a’所代表的灰色区域表示在球面上相邻但在全景投影展开图像上不相邻的面边界区域对,其中a的取值可以为1、2、3、4、5、6、7以及8中的任一个。图3中的白色区域表示全景投影展开图像上的非面边界区域,而黑色区域表示在球面上相邻且在全景投影展开图像上仍然相邻的面边界区域。
传统的图像或视频编解码方法是将整幅图像划分成像素块来进行处理的,由于像素块之间在预测、变换、量化等过程中存在不一致性,像素块边界处存在着不连续性,因此要对像素块边界进行去块滤波处理。传统的去块滤波方法中,整幅图像的去块滤波处理的方向是一致的,并且去块滤波处理是像素块为单位的。
传统的去块滤波方法为二维分离的滤波,为了在编解码端匹配,因此需要指定第一维和第二维滤波处理的方向,并对像素块依次进行第一维和第二维滤波处理。例如先对像素块在图像上的边界方向(如竖直方向)进行第一维滤波处理,再对像素块在图像上的另一边界方向(如水平方向)进行第二维滤波处理,其中像素块的边界方向和滤波处理方向相互垂直,即对竖直边界进行水平方向的滤波处理,第一维滤波处理方向和第二维滤波处理方向也相互垂直。如图4所示,像素块A和B之间的边界方向为竖直方向,因此先对像素块A、B之间的竖直边界进行第一维滤波处理,即先进行水平方向的滤波,像素块A和C之间为水平边界,将第一维滤波处理后的结果再进行第 二维滤波处理,即再进行竖直方向的滤波。在HEVC(High Efficiency Video Coding)以及H.264中,在对像素块边界进行去块滤波操作时先对竖直边界进行水平方向的滤波再对水平边界进行竖直方向的滤波,即先水平滤波再垂直滤波。
传统的编解码方法中,编码或解码顺序是从上到下从左到右的,如图5所示,对于像素块A、B、C以及D来说,先编码或解码像素块A,再编码或解码像素块B,接着编码或解码像素块C,最后编码或解码像素块D。因此在图3中在球面上相邻但在全景投影展开图像上不相邻的面边界区域对a和a’中,a’是较后编码或解码处理的面边界区域,其中a取1、2、3、4、5、6、7以及8中的任意一个。
对全景图像展开并进行压缩编码时,压缩编码存在失真,在编解码图像中的不同位置的失真不同。通过解码图像生成球面图像时,由于编解码上不同位置的失真不同,球面分割线两个的像素在编解码图像上的失真不同,所以在通过解码图像生成球面图像时会出现伪像,可以通过对编解码图像中球面分割线处进行滤波来减弱甚至消除伪像现象。
目前图像编解码器中的滤波技术都是对在编解码图像中相邻的像素块进行滤波操作的,对于全景图像球面分割线两侧的像素块不进行滤波。为了去除全景图像观看图像中的伪像,需要对全景视频球面分割线两侧的像素块设计特殊的滤波方法,由于全景视频球面分割线两侧的像素块在解码图像上存在不对齐情况,针对这一现象需要设计相应的对齐方法。
发明内容
本发明的目的是解决在全景球面上相邻、但在投影展开图像上不相邻的面边界区域对交界处的不连续性。通过指定面边界区域对中的一个特定面边界来确定面边界区域对第一维和第二维滤波处理的方向,确保编解码端对面边界区域对进行滤波处理时可以实现匹配。
本发明第一目的在于提供一种全景图像滤波方法,包括:
根据全景投影展开图像的格式信息,确定在全景投影展开图像中存在的在全景球面上相邻、但在投影展开图像上不相邻的面边界区域对的位置;
按照所述面边界区域对中一个特定的面边界区域在全景投影展开图像上的方向,确定面边界区域对中第一维和第二维滤波处理的方向,所述面边界区域对确定第一维和第二维滤波处理方向的方法与非面边界区域在全景投影展开图像上确定第一维和第二维滤波处理的方向一致;
对所述面边界区域对依次进行第一维和第二维滤波处理。
进一步的,所述特定的面边界区域为所述面边界区域对中在全景投影展开图像上较后编码或解码处理的面边界区域。
进一步的,对于同属于两个面边界区域的共同子区域,所述共同子区域滤波后的像素由所述两个面边界区域中在全景展开投影图像上较后编码或解码处理的面边界区域所对应的子区域滤波后的像素来确定。
本发明第二目的在于提供一种全景图像滤波装置,包括以下模块:
待滤波面边界区域对位置导出模块:该模块的输入为全景投影展开图像的格式信息,根据格式信息确定在全景投影展开图像中存在的在全景球面上相邻、但在投影展 开图像上不相邻的面边界区域对的位置,该模块的输出为所述待滤波面边界区域对在全景投影展开图像中的位置;
滤波处理方向确定模块:该模块输入待滤波面边界区域对及其在全景投影展开图像中的位置,按照所述面边界区域对中一个特定的面边界区域在全景投影展开图像上的方向,确定面边界区域对中第一维和第二维滤波处理的方向,所述面边界区域对确定第一维和第二维滤波处理方向的方法与非面边界区域在全景投影展开图像上确定第一维和第二维滤波处理的方向一致;该模块输出待滤波面边界区域对的第一维和第二维滤波处理的方向;
滤波处理模块:输入为待滤波面边界区域对的第一维和第二维滤波处理的方向,对所述面边界区域对依次进行第一维和第二维滤波处理。
进一步的,所述滤波处理方向确定模块中的一个特定面边界区域为所述面边界区域对中在全景投影展开图像上较后编码或解码处理的面边界区域。
进一步的,所述滤波处理模块中,对于同属于两个面边界区域的共同子区域,所述共同子区域滤波后的像素由所述两个面边界区域中在全景展开投影图像上较后编码/解码处理的面边界区域所对应的子区域滤波后的像素来确定。
本发明第三目的是提出一种全景图像像素块滤波方法,包括:根据紧邻全景图像球面分割线一侧的待滤波像素块所在图像对应的投影格式信息,对所述待滤波像素块中的标称点沿所述球面分割线方向的位置进行伸缩变换,确定所述标称点在紧邻所述球面分割线另一侧的对应标称点的位置;由所述对应标称点确定所述待滤波像素块在紧邻所述球面分割线另一侧的对应像素块;对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
作为优选,所述待滤波像素块的对应像素块中对应标称点位置的确定方法为以下情况之一:
1)对所述对应像素块中对应标称点的位置取有限精度得到整像素位置;
2)对所述对应像素块中对应标称点的位置取有限精度得到分像素位置。
作为优选,所述待滤波像素块的对应像素块中其余非对应标称点的位置的确定方法为:以所述对应像素块中对应标称点为基准,按整像素间隔确定所述对应像素块中非对应标称点的位置。
作为优选,本滤波方法还包括获取所述待滤波像素块及其对应像素块的编码信息,根据所述待滤波像素块及其对应像素块的编码信息确定滤波方法。
作为优选,所述待滤波像素块对应像素块的编码信息的获取方法为以下情况之一:
1)所述编码信息来自所述对应像素块中对应标称点所在的编码单元;
2)所述编码信息来自包含所述对应像素块中像素数目最多的编码单元。
本发明第四目的在于提出一种全景图像像素块滤波方法,包括:
对紧邻全景图像球面分割线一侧的待滤波像素块,根据所述待滤波像素块所在图像对应的投影格式信息,确定所述待滤波像素块在紧邻所述球面分割线另一侧的对应像素块的位置;根据所述待滤波像素块及其对应像素块的编码信息确定滤波方法;其中所述待滤波像素块及其对应像素块中的编码信息中的运动信息始终认为相同;对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
本发明第五目的是提出一种全景图像像素块滤波装置,其包括以下模块:
对应标称点位置导出模块:该模块的输入为紧邻全景图像球面分割线一侧的待滤波像素块中标称点的位置信息以及所述待滤波像素块所在图像对应的投影格式信息;输出为所述标称点在全景图像中对应标称点的位置;根据紧邻全景图像球面分割线一侧的待滤波像素块所在图像对应的投影格式信息,对所述待滤波像素块中的标称点沿所述球面分割线方向的位置进行伸缩变换,确定所述标称点在紧邻所述球面分割线另一侧的对应标称点的位置;
对应像素块位置导出模块:该模块输入为对应标称点的位置,输出为待滤波像素块在全景图像中对应像素块的位置;该模块用于根据对应标称点的位置确定对应像素块中其余非对应标称点的位置;
像素滤波模块:该模块输入为待滤波像素块位置及其对应像素块的位置,输出为待滤波像素块滤波后的像素;该模块用于对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
作为优选,所述对应标称点位置导出模块中还包括对应标称点位置取整模块,所述对应标称点位置取整模块中对应标称点位置的取整方法为以下情况之一:
1)对所述对应像素块中对应标称点的位置取有限精度得到整像素位置;
2)对所述对应像素块中对应标称点的位置取有限精度得到分像素位置。
作为优选,所述对应像素块位置导出模块中还包括非对应标称点位置导出模块,所述非对应标称点位置导出模块中非标称点位置的确定方法为:以对应像素块中对应标称点为基准,按整像素间隔确定所述非对应标称点的位置。
作为优选,所述像素滤波模块中还包括以下子模块:
编码信息获取模块,用于获取所述待滤波像素块及其对应像素块的编码信息;
滤波方法确定模块,用于根据所述待滤波像素块及其对应像素块的编码信息和像素特性确定滤波方法。
作为优选,所述待滤波像素块对应像素块的编码信息的获取方法为以下情况之一:
1)所述编码信息来自所述对应像素块中对应标称点所在的编码单元;
2)所述编码信息来自包含所述对应像素块中像素数目最多的编码单元。
作为优选,还包括以下模块:
对应像素块位置导出模块:该模块输入为紧邻全景图像球面分割线一侧的待滤波像素块的位置信息以及所述待滤波像素块所在图像对应的投影格式信息,输出为待滤波像素块的对应像素块位置;
编码信息获取模块,用于获取所述待滤波像素块及其对应像素块的编码信息,其中所述待滤波像素块及其对应像素块的编码信息中的运动信息始终认为相同;
像素滤波模块,用于对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
本发明第六目的是提出一种全景图像像素块滤波装置,其包括以下模块:
对应像素块位置导出模块:该模块输入为紧邻全景图像球面分割线一侧的待滤波像素块的位置信息以及所述待滤波像素块所在图像对应的投影格式信息,输出为待滤波像素块的对应像素块位置;
滤波方法确定模块,用于确定滤波方法,输入为待滤波像素块及其对应像素块的位置信息,根据所述待滤波像素块及其对应像素块的编码信息确定滤波方法,输出为 该模块确定的滤波方法。
作为优选,所述确定滤波方法为以下之一种:
1)认为所述待滤波像素块及其对应像素块的编码信息中的运动信息相同;
2)不使用所述待滤波像素块及其对应像素块的编码信息中的运动信息。
像素滤波模块,用于对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
本发明的有益效果可以减少全景图像的编码伪像和面边界的不连续性,提高了全景图像编码后的主观质量,同时规定第一维和第二维滤波处理的方向后能够确保在对面边界区域进行滤波处理时可以实现编解码端的匹配。
附图说明
结合附图,可从下面举出的实施例来对本发明的原理进行解释。
此处所说明的附图用来提供对本发明的进一步理解,构成本申请的一部分,本发明的实施例仅用来解释本发明,只是其中的一些特例,本发明的适用范围并不仅限于这些实施例。在附图中:
图1将球面通过透视投影的方式映射到球的外接立方体上,并将外接立方体六个面展开;
图2表示将立方体六个面展开后通过旋转紧凑排布成3x2的格式;
图3表示全景投影展开图像上面边界区域和非面边界区域的分布;
图4为第一维滤波处理和第二维滤波处理的方向及处理顺序;
图5为传统编解码技术中一个图像内像素块的编码或解码顺序;
图6为本发明一个实施例中在球面上相邻但在全景投影展开图像上不相邻的面边界区域对;
图7为本发明一个实施例中在球面上相邻但在全景投影展开图像上不相邻的面边界区域对中的待滤波像素块;
图8为本发明一个实施例中在球面上相邻但在全景投影展开图像上不相邻的面边界区域对2和2’中的待滤波像素块的滤波方向及顺序;
图9为本发明一个实施例中在球面上相邻但在全景投影展开图像上不相邻的面边界区域对1和1’中的待滤波像素块的对齐方式;
图10为本发明一个实施例中在在球面上相邻但在全景投影展开图像上不相邻的面边界区域对3和3’中的待滤波像素块的对齐方式;
图11为本发明一个实施例中在球面上相邻但在全景投影展开图像上不相邻的面边界区域对4和4’中的待滤波像素块;
图12为本发明一个实施例中在球面上相邻但在全景投影展开图像上不相邻的面边界区域对4和4’中的待滤波像素块的滤波方向及顺序;
图13为本发明一个实施例中在球面上相邻但在全景投影展开图像上不相邻的面边界区域对5和5’中的待滤波像素块;
图14为本发明一个实施例中在球面上相邻但在全景投影展开图像上不相邻的面边界区域对5和5’中的待滤波像素块的滤波方向及顺序;
图15为本发明一个实施例中在球面上相邻但在全景投影展开图像上不相邻的面边 界区域对7和7’中的待滤波像素块;
图16为本发明一个实施例中在球面上相邻但在全景投影展开图像上不相邻的面边界区域对7和7’中的待滤波像素块的滤波方向及顺序;
图17为本发明一个实施例中的装置图;
图18为本发明一个实施例中同属于两个面边界区域的共同子区域示意图;
图19为对共同子区域进行滤波处理的流程。
图20将六面体六个面展开并紧凑排布后的位置关系示意图;
图21为本发明一个实施例中待滤波像素块及其对应像素块在全景图像展开图像中的位置示意图;
图22为本发明一个实施例中待滤波像素块中的一个标称点的位置示意图;
图23为本发明一个实施例中对应标称点在对应像素块中的位置示意图;
图24为本发明一个实施例中对应标称点在对应像素块中的位置示意图;
图25为本发明一个实施例中待滤波像素块中的一个标称点的位置示意图;
图26为本发明一个实施例中对应标称点在对应像素块中的位置以及对应块所述编码单元示意图;
图27为本发明一个实施例中待滤波像素块及其对应像素块在全景图像展开为六面体图像中的位置示意图;
图28为本发明一个实施例中待滤波像素块及其对应像素块在全景图像展开为经纬图中的位置示意图;
图29为本发明一个实施例中全景图像像素块滤波装置示意图;
图30为本发明一个实施例中全景图像像素块滤波装置示意图;
图31为本发明一个实施例中对应标称点位置导出模块中子模块示意图;
图32为本发明一个实施例中对应像素块位置导出模块中子模块示意图;
图33为本发明一个实施例中像素滤波模块中子模块示意图;
图34为编码导致的全景图像viewport中的伪像示意图;
图35为经过滤波后的全景图像viewport;
图36为本发明一个实施例中边界强度的判决流程;
图37为HEVC去块滤波中边界强度的判决流程。
具体实施方式
实施例1
本发明实施例提供了一种全景图像滤波方法。
本实施例提供的全景图像滤波方法包括:
根据全景投影展开图像的格式信息,确定在全景投影展开图像中存在的在全景球面上相邻、但在投影展开图像上不相邻的面边界区域对的位置;
按照所述面边界区域对中一个特定的面边界区域在全景投影展开图像上的方向,确定面边界区域对中第一维和第二维滤波处理的方向,所述面边界区域对确定第一维和第二维滤波处理方向的方法与非面边界区域在全景投影展开图像上确定第一维和第二维滤波处理的方向一致;
对所述面边界区域对依次进行第一维和第二维滤波处理。
在本实施例中所述特定的面边界区域为面边界区域对中在全景投影展开图像上后 编码或解码处理的面边界区域;
如图6所示立方体格式的全景投影展开图像,且排布成3x2的矩形图像,因此可以很容易确定图6所示的全景投影展开图像上的面边界区域a和a’为全景图像中在全景球面上相邻、但在全景投影展开图像上不相邻的面边界区域对,其中a取1、2、3、4、5、6、7以及8中的任意一个。由于预测、变换、量化等编解码是在全景投影展开图像上进行的,那些在全景球面上相邻、但在全景投影展开图像上不相邻的像素块之间会出现特殊的不连续性,因此需要对这些像素块进行去块滤波以平滑其不连续性。如图6所示,面边界区域2中的像素块b和面边界区域2’中的像素块b’即为在全景球面上相邻但在全景投影展开图像上不相邻的像素块,因此需要对像素块b和b’进行去块滤波以平滑像素块间的不连续性。
在图6面边界区域对2和2’中,面边界区域2’为较后编码或解码处理的区域,且面边界区域2’在全景投影展开图像上的方向为竖直方向,在本实施例中设置非面边界区域第一维滤波处理的方向为水平方向,第二维滤波处理的方向为竖直方向,但本发明也不仅仅限于此设置,例如也可以设置成第一维滤波处理的方向为竖直方向,第二维滤波处理的方向为水平方向。当非面边界区域第一维滤波处理的方向为水平方向时,面边界区域对2和2’的第一维滤波处理方向也为水平方向,对面边界区域对2和2’中的像素块依次进行第一维和第二维滤波处理。
如图7所示,面边界区域2中有像素块d和e,面边界区域2’中有像素块d’和e’,对像素块d、e、d’以及e’进行滤波处理时,先将像素块d和e逆时针旋转90°后和像素块d’以及e’对齐,然后对像素块依次进行第一维和第二维滤波处理。如图8所示,首先对像素块d和d’,e和e’进行水平方向的去块滤波处理,得到滤波处理后的结果,再对像素块d’和e’,d和e进行竖直方向的去块滤波处理。按照本实施例采用的去块滤波处理,面边界区域2中的像素块的第一维和第二维滤波处理顺序和非面边界区域中的像素块的第一维和第二维滤波处理顺序不同。
对于图7中的其他编号的面边界区域对也按照本实施例中所描述的方法确定滤波方向对面边界区域对的边界进行滤波。例如面边界区域1和1’中,面边界区域1’在全景投影展开图像上的方向为水平方向,则面边界区域对1和1’中的像素块滤波顺序为先进行竖直边界的滤波再进行水平边界的滤波,面边界区域1中的像素块不需要进行任何旋转或转置直接和面边界区域1’中的像素块进行去块滤波处理,如图9所示,面边界区域对1中的像素第一维和第二维滤波处理顺序没有改变;面边界区域3和3’中,3’在全景投影展开图像上的方向为水平方向。则面边界区域对3和3’中的像素块滤波顺序为先进行竖直边界的滤波再进行水平边界的滤波,需要将面边界区域3中的像素块旋转180°之后和面边界区域3’进行去块滤波处理,如图10所示,面边界区域对3中的像素第一维和第二维滤波处理顺序没有改变。
对图6中的其他在球面上相邻但在全景投影展开图像上不相邻的面边界区域对中的像素块也按照本实施例中所述的方法处理。
实施例2
本发明实施例提供了一种全景图像滤波方法。
本实施例提供的全景图像滤波方法包括:
根据全景投影展开图像的格式信息,确定在全景投影展开图像中存在的在全景球 面上相邻、但在投影展开图像上不相邻的面边界区域对的位置;
按照所述面边界区域对中一个特定的面边界区域在全景投影展开图像上的方向,确定面边界区域对中第一维和第二维滤波处理的方向,所述面边界区域对确定第一维和第二维滤波处理方向的方法与非面边界区域在全景投影展开图像上确定第一维和第二维滤波处理的方向一致;
对所述面边界区域对依次进行第一维和第二维滤波处理。
在本实施例中所述特定的面边界区域为面边界区域对中在全景投影展开图像上先编码或解码处理的面边界区域;
在图6面边界区域对4和4’中,面边界区域4为先编码或解码处理的区域,且面边界区域4在全景投影展开图像上的方向为竖直方向,在本实施例中设置非面边界区域第一维滤波处理的方向为水平方向,第二维滤波处理的方向为竖直方向,但本发明也不仅仅限于此设置,例如也可以设置成第一维滤波处理的方向为竖直方向,第二维滤波处理的方向为水平方向。当非面边界区域第一维滤波处理的方向为水平方向时,面边界区域对4和4’的第一维滤波处理方向也为水平方向,对面边界区域对4和4’中的像素块依次进行第一维和第二维滤波处理。
如图11所示,面边界区域4中有像素块b和c,面边界区域4’中有像素块b’和c’,对像素块b、c、b’以及c’进行滤波处理时,先将像素块b’和c’逆时针旋转90°后和像素块b以及c对齐,然后对像素块依次进行第一维和第二维滤波处理。如图12所示,首先对像素块b’和b,c和c’进行水平方向的去块滤波处理,得到滤波处理后的结果,再对像素块b’和c’,b和c进行竖直方向的去块滤波处理。按照本实施例采用的去块滤波处理,面边界区域4中的像素块的第一维和第二维滤波处理顺序和非面边界区域中的像素块的第一维和第二维滤波处理顺序不同。
对图6中的其他在球面上相邻但在全景投影展开图像上不相邻的面边界区域对中的像素块也按照本实施例中所述的方法处理。
实施例3
本发明实施例提供了一种全景图像滤波方法。
本实施例提供的全景图像滤波方法包括:
根据全景投影展开图像的格式信息,确定在全景投影展开图像中存在的在全景球面上相邻、但在投影展开图像上不相邻的面边界区域对的位置;
按照所述面边界区域对中一个特定的面边界区域在全景投影展开图像上的方向,确定面边界区域对中第一维和第二维滤波处理的方向,所述面边界区域对确定第一维和第二维滤波处理方向的方法与非面边界区域在全景投影展开图像上确定第一维和第二维滤波处理的方向一致;
对所述面边界区域对依次进行第一维和第二维滤波处理。
在本实施例中若所述面边界区域对中两个面边界区域的方向不同,所述特定的面边界区域为面边界区域对中水平方向的面边界区域。
在图6面边界区域对5和5’中,面边界区域5和面边界区域5’在全景投影展开图像上的方向不同。面边界区域5’为水平方向的面边界区域,在本实施例中设置非面边界区域第一维滤波处理的方向为竖直方向,第二维滤波处理的方向为水平方向,但本发明也不仅仅限于此设置,例如也可以设置成第一维滤波处理的方向为水平方向, 第二维滤波处理的方向为竖直方向。当非面边界区域第一维滤波处理的方向为竖直方向时,面边界区域对5和5’的第一维滤波处理方向也为竖直方向,对面边界区域对5和5’中的像素块依次进行第一维和第二维滤波处理。
如图13所示,面边界区域5中有像素块b和c,面边界区域5’中有像素块b’和c’,对像素块b、c、b’以及c’进行滤波处理时,先将像素块b和c顺时针旋转90°后和像素块b’以及c’对齐,然后对像素块依次进行第一维和第二维滤波处理。如图14所示,首先对像素块b’和b,c’和c进行竖直方向的去块滤波处理,得到滤波处理后的结果,再对像素块b和c,b’和c’进行水平方向的去块滤波处理。按照本实施例采用的去块滤波处理,面边界区域5中的像素块的第一维和第二维滤波处理顺序和非面边界区域中的像素块的第一维和第二维滤波处理顺序相同。
对图6中的其他在球面上相邻但在全景投影展开图像上不相邻的面边界区域对中的像素块也按照本实施例中所述的方法处理。
实施例4
本发明实施例提供了一种全景图像滤波方法。
本实施例提供的全景图像滤波方法包括:
根据全景投影展开图像的格式信息,确定在全景投影展开图像中存在的在全景球面上相邻、但在投影展开图像上不相邻的面边界区域对的位置;
按照所述面边界区域对中一个特定的面边界区域在全景投影展开图像上的方向,确定面边界区域对中第一维和第二维滤波处理的方向,所述面边界区域对确定第一维和第二维滤波处理方向的方法与非面边界区域在全景投影展开图像上确定第一维和第二维滤波处理的方向一致;
对所述面边界区域对依次进行第一维和第二维滤波处理。
在本实施例中若所述面边界区域对中两个面边界区域的方向不同,所述特定的面边界区域为面边界区域对中竖直方向的面边界区域。
在图6面边界区域对7和7’中,面边界区域7和面边界区域7’在全景投影展开图像上的方向不同。面边界区域7’为竖直方向的面边界区域,在本实施例中设置非面边界区域第一维滤波处理的方向为水平方向,第二维滤波处理的方向为竖直方向,但本发明也不仅仅限于此设置,例如也可以设置成第一维滤波处理的方向为竖直方向,第二维滤波处理的方向为水平方向。当非面边界区域第一维滤波处理的方向为水平方向时,面边界区域对7和5’的第一维滤波处理方向也为水平方向,对面边界区域对7和7’中的像素块依次进行第一维和第二维滤波处理。
如图15所示,面边界区域7中有像素块b和c,面边界区域7’中有像素块b’和c’,对像素块b、c、b’以及c’进行滤波处理时,先将像素块b和c顺时针旋转90°后和像素块b’以及c’对齐,然后对像素块依次进行第一维和第二维滤波处理。如图16所示,首先对像素块b’和b,c’和c进行水平方向的去块滤波处理,得到滤波处理后的结果,再对像素块b和c,b’和c’进行竖直方向的去块滤波处理。按照本实施例采用的去块滤波处理,面边界区域7中的像素块的第一维和第二维滤波处理顺序和非面边界区域中的像素块的第一维和第二维滤波处理顺序不同。
对图6中的其他在球面上相邻但在全景投影展开图像上不相邻的面边界区域对中的像素块也按照本实施例中所述的方法处理。
实施例5
本发明实施例提供了一种全景图像滤波装置。
本实施例提供的全景图像滤波装置包括:
待滤波面边界区域对位置导出模块:该模块的输入为全景投影展开图像的格式信息,根据格式信息确定在全景投影展开图像中存在的在全景球面上相邻、但在投影展开图像上不相邻的面边界区域对的位置,该模块的输出为所述待滤波面边界区域对在全景投影展开图像中的位置;
滤波处理方向确定模块:该模块输入待滤波面边界区域对及其在全景投影展开图像中的位置,按照所述面边界区域对中一个特定的面边界区域在全景投影展开图像上的方向,确定面边界区域对中第一维和第二维滤波处理的方向,所述面边界区域对确定第一维和第二维滤波处理方向的方法与非面边界区域在全景投影展开图像上确定第一维和第二维滤波处理的方向一致;该模块输出待滤波面边界区域对的第一维和第二维滤波处理的方向;
滤波处理模块:输入为待滤波面边界区域对的第一维和第二维滤波处理的方向,对所述面边界区域对依次进行第一维和第二维滤波处理。
图17给出了本实施例中全景图像滤波装置示意图。其中在待滤波面边界区域对位置导出模块中,输入的全景投影展开图像的格式信息确定了全景投影展开图像的展开方式,如立方体格式、EAC、PAU等,本实施例以立方体格式为例,如图2所示,全景投影展开图像被投影到立方体的六个面上且将6个面排布成3x2的矩形图像。在图6所示的相同数字对应的区域即为待滤波面边界区域对。
在所述滤波处理方向确定模块中,需要指定面边界区域对中的一个面边界区域,根据该面边界区域在全景投影展开图像中的方向来指定所述面边界区域对的第一维和第二维滤波处理方向。指定的面边界区域为以下4种情况之一:
a)所述指定的面边界区域为面边界区域对中在全景投影展开图像上先编码或解码处理的面边界区域;
b)所述指定的面边界区域为面边界区域对中在全景投影展开图像上后编码或解码处理的面边界区域;
c)若所述面边界区域对中两个面边界区域的方向不同,所述指定的面边界区域为面边界区域对中水平方向的面边界区域;
d)若所述面边界区域对中两个面边界区域的方向不同,所述指定的面边界区域为面边界区域对中竖直方向的面边界区域;
然后滤波处理方向确定模块根据所指定的面边界区域的方向指定面边界区域对的第一维和第二维滤波处理方向。
滤波处理模块对所述面边界区域对依次进行第一维和第二维滤波处理。在所述滤波处理模块中,对于同属于两个面边界区域的共同子区域,所述共同子区域滤波后的像素由所述两个面边界区域中在全景展开投影图像上按照编解码顺序较后编码/解码处理的面边界区域所对应的子区域滤波后的像素来确定。
对图6中的其他在球面上相邻但在全景投影展开图像上不相邻的面边界区域对中的像素块也按照本实施例中所述的装置处理。
实施例6
本发明实施例提供了一种全景图像滤波方法。
本实施例提供的全景图像滤波方法包括:
所述共同子区域滤波后的像素由所述两个面边界区域中在全景展开投影图像上按照编解码顺序较后编码/解码处理的面边界区域所对应的子区域滤波后的像素来确定。
在本实施例中,如图18所示,像素块b同属于面边界区域8和面边界区域5,像素块b即为所述全景图像滤波方法中的共同子区域。面边界区域8在球面上相邻但在全景投影展开图像上不相邻的面边界区域为8’;面边界区域5在球面上相邻但在全景投影展开图像上不相邻的面边界区域为5’。像素块b在面边界区域8’中的球面相邻像素块为b 1’,像素块b在面边界区域5’中的球面相邻像素块为b 2’。在对像素块b、b 1’以及b 2’进行滤波处理时,将面边界区域5中的像素块b顺时针旋转90°后放在面边界区域5’中像素块b 2’的上方,如图19所示。
在本实施例中指定根据面边界区域对中后编码或解码处理的面边界区域在全景投影展开图像上的方向确定面边界区域对第一维和第二维滤波处理的方向。本实施例中设非面边界区域的第一维滤波处理方向为竖直方向,则对像素块b、b 1’以及b 2’进行滤波处理时,先对像素块b和b 1’进行竖直方向的去块滤波处理,对像素块b和b 2’进行竖直方向的去块滤波处理,然后对面边界区域8中的像素块b和面边界区域5中的像素块b进行水平方向的滤波。全景投影展开图像上像素块b滤波后的结果由面边界区域5中的像素块b确定。因为面边界区域5相对于面边界区域8按照编解码顺序较后处理。
若像素块间是串行依次进行去块滤波处理的,则在全景图像上靠下和靠右的像素块会较后处理,即图19中面边界区域5相对于面边界区域8较后去块滤波处理,若像素块间是并行进行去块滤波处理的,则共同子区域滤波后的像素仍由按照串行处理时较后处理的面边界区域确定,这样保证了串行处理和并行处理的一致性。
本发明通过确定面边界区域对的第一维和第二维滤波处理的方向以及处理顺序,使得编解码端在对面边界区域对进行滤波处理时可以匹配。
实施例7
本发明实施例提供了一种全景图像像素块滤波方法。
本实施例提供的全景图像像素块滤波方法包括:
根据紧邻全景图像球面分割线一侧的待滤波像素块所在图像对应的投影格式信息,对所述待滤波像素块中的标称点沿所述球面分割线方向的位置进行伸缩变换,确定所述标称点在紧邻所述球面分割线另一侧的对应标称点的位置;由所述对应标称点确定所述待滤波像素块在紧邻所述球面分割线另一侧的对应像素块;对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
本实施例中所述待滤波像素块所在的图像为3x2的六面体格式,如图20所示。图21给出了一个待滤波像素块B及其在全景图像球面分割线另一侧的对应像素块B’。假设图像在X-Y坐标系下采样率相同,由所述格式信息可以得到图像在图21中的X’-Y’坐标系下采样率不同,且有以下关系:
x′=f(x)
y′=g(y)
其中x∈(-1,1),y∈(-1,1),x′∈(-1,1),y′∈(-1,1)。
如图22所示所述标称点为待滤波像素块B中最靠近面边界一行像素的几何中点b,对该点的横坐标进行伸缩变换后得到像素块B’中对应标称点b’的纵坐标。伸缩变换方法如下:
x=f -1(x′)
y′=g(x)
其中f -1(x′)表示f(x)的反函数,在求y′对应公式中的y为±1。
图23给出了对应像素块B’中的对应标称点b’,滤波所需的编码信息如量化步长、预测模式(帧内预测或帧间预测)等信息来自标称点b’所在的编码单元,同时将b’的纵坐标取有限精度得到整像素精度,非对应标称点的像素以标称点b’位置为基准按照整像素间隔获取像素,根据获取待滤波像素块及其对应像素块的编码信息和像素特性确定滤波方法,对待滤波像素块及其对应像素块滤波,得到待滤波像素块滤波后的像素。
实施例8
本发明实施例提供了一种全景图像像素块滤波方法。
本实施例提供的全景图像像素块滤波方法包括:
根据紧邻全景图像球面分割线一侧的待滤波像素块所在图像对应的投影格式信息,对所述待滤波像素块中的标称点沿所述球面分割线方向的位置进行伸缩变换,确定所述标称点在紧邻所述球面分割线另一侧的对应标称点的位置;由所述对应标称点确定所述待滤波像素块在紧邻所述球面分割线另一侧的对应像素块;对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
本实施例中所述待滤波像素块所在的图像为3x2的六面体格式,如图20所示。图21给出了一个待滤波像素块B及其在全景图像球面分割线另一侧的对应像素块B’。假设图像在X-Y坐标系下采样率相同,由所述格式信息可以得到图像在图21中的X’-Y’坐标系下采样率不同,且有以下关系:
x′=f(x)
y′=g(y)
其中x∈(-1,1),y∈(-1,1),x′∈(-1,1),y′∈(-1,1)。
如图22所示所述标称点为待滤波像素块B中最靠近面边界一行像素的几何中点b,对该点的横坐标进行伸缩变换后得到像素块B’中对应标称点b’的纵坐标。伸缩变换方法如下:
x=f -1(x′)
y′=g(x)
其中f -1(x′)表示f(x)的反函数。
图24给出了对应像素块B’中的对应标称点b’,滤波所需的编码信息如量化步长、预测模式(帧内预测或帧间预测)等信息来自标称点b’所在的编码单元,对B’中的对应标称点b’的纵坐标取有限精度得到分像素精度,例如1/2像素精度、1/4像素精度、1/8像素精度、1/16像素精度等。其余非对应标称点所在位置的像素以b’为基准按照整像素间隔获取,非整像素位置的像素由周围整像素位置的像素插值得到,根据获取待滤波像素块及其对应像素块的编码信息和像素特性确定滤波方法。本实施例中所述滤波特指去块滤波,确定滤波方法指确定去块滤波时的滤波系数。根据确定 的滤波系数对待滤波像素块及其对应像素块滤波,得到待滤波像素块滤波后的像素。
实施例9
本发明实施例提供了一种全景图像像素块滤波方法。
本实施例提供的全景图像像素块滤波方法包括:
根据紧邻全景图像球面分割线一侧的待滤波像素块所在图像对应的投影格式信息,对所述待滤波像素块中的标称点沿所述球面分割线方向的位置进行伸缩变换,确定所述标称点在紧邻所述球面分割线另一侧的对应标称点的位置;由所述对应标称点确定所述待滤波像素块在紧邻所述球面分割线另一侧的对应像素块;对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
本实施例中所述待滤波像素块所在的图像为3x2的六面体格式,如图20所示。图21给出了一个待滤波像素块B及其在全景图像球面分割线另一侧的对应像素块B’。假设图像在X-Y坐标系下采样率相同,由所述格式信息可以得到图像在图21中的X’-Y’坐标系下采样率不同,且有以下关系:
x′=f(x)
y′=g(y)
其中x∈(-1,1),y∈(-1,1),x′∈(-1,1),y′∈(-1,1)。
如图25所示所述标称点为待滤波像素块B中最靠近面边界一行像素左侧像素点b,对该点的横坐标进行伸缩变换后得到像素块B’中对应标称点b’的纵坐标。伸缩变换方法如下:
x=f -1(x′)
y′=g(x)
其中f -1(x′)表示f(x)的反函数。
图26给出了对应像素块B’中的对应标称点b’,,由于编码单元C包含B’中的像素数目最多,所以滤波所需的编码信息如量化步长、预测模式(帧内预测或帧间预测)等信息来编码单元C,同时将标称点b’的纵坐标取有限精度得到整像素精度,非对应标称点的像素以标称点b’位置为基准按照整像素间隔获取像素,根据获取待滤波像素块及其对应像素块的编码信息和像素特性确定滤波方法,对待滤波像素块及其对应像素块滤波,得到待滤波像素块滤波后的像素。
实施例10
本发明实施例提供了一种全景图像像素块滤波方法。
本实施例提供的全景图像像素块滤波方法包括:
根据紧邻全景图像球面分割线一侧的待滤波像素块所在图像对应的投影格式信息,对所述待滤波像素块中的标称点沿所述球面分割线方向的位置进行伸缩变换,确定所述标称点在紧邻所述球面分割线另一侧的对应标称点的位置;由所述对应标称点确定所述待滤波像素块在紧邻所述球面分割线另一侧的对应像素块;对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
本实施例中所述待滤波像素块所在的图像为3x2的六面体格式,如图20所示。图21给出了一个待滤波像素块B及其在全景图像球面分割线另一侧的对应像素块B’。假设图像在X-Y坐标系下采样率相同,由所述格式信息可以得到图像在图21中的X’-Y’坐标系下采样率不同,且有以下关系:
x′=f(x)
y′=g(y)
其中x∈(-1,1),y∈(-1,1),x′∈(-1,1),y′∈(-1,1)。
如图25所示所述标称点为待滤波像素块B中最靠近面边界一行像素左侧像素点b,对该点的横坐标进行伸缩变换后得到像素块B’中对应标称点b’的纵坐标。伸缩变换方法如下:
x=f -1(x′)
y′=g(x)
其中f -1(x′)表示f(x)的反函数。
图26给出了待滤波像素块B在全景图像上的对应像素B’中的对应标称点b’,由于编码单元C包含B’中的像素数目最多,所以滤波所需的编码信息如量化步长、预测模式(帧内预测或帧间预测)等信息来编码单元C,同时将标称点b’的纵坐标取有限精度得到分像素精度,例如1/2像素精度、1/4像素精度、1/8像素精度、1/16像素精度等。非对应标称点的像素以标称点b’位置为基准按照整像素间隔获取像素,根据获取待滤波像素块及其对应像素块的编码信息和像素特性确定滤波方法,对待滤波像素块及其对应像素块滤波,得到待滤波像素块及其对应像素块滤波后的像素。
实施例11
本发明实施例提供了一种全景图像像素块滤波方法。
本实施例提供的全景图像像素块滤波方法包括:
根据紧邻全景图像球面分割线一侧的待滤波像素块所在图像对应的投影格式信息,对所述待滤波像素块中的标称点沿所述球面分割线方向的位置进行伸缩变换,确定所述标称点在紧邻所述球面分割线另一侧的对应标称点的位置;由所述对应标称点确定所述待滤波像素块在紧邻所述球面分割线另一侧的对应像素块;对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
本实施例中所述待滤波像素块所在的图像为3x2的六面体格式,如图20所示。图21给出了一个待滤波像素块B及其在全景图像球面分割线另一侧的对应像素块B’。假设图像在X-Y坐标系下采样率相同,由所述格式信息可以得到图像在图21中的X’-Y’坐标系下采样率不同,且有以下关系:
x′=f(x)
y′=g(y)
其中x∈(-1,1),y∈(-1,1),x′∈(-1,1),y′∈(-1,1)。
如图22所示所述标称点为待滤波像素块B中最靠近面边界一行像素的几何中点b,根据该点横坐标计算出像素块B’中对应标称点b’的纵坐标。计算方法如下:
x=f -1(x′)
y′=g(x)
其中f -1(x′)表示f(x)的反函数。
图23给出了对应像素块B’中的对应标称点b’,对B’中的对应标称点b’的纵坐标取有限精度得到分像素精度,例如1/2像素精度、1/4像素精度、1/8像素精度、1/16像素精度等。其余非对应标称点所在位置的像素以b’为基准按照整像素间隔获取,非整像素位置的像素由周围整像素位置的像素插值得到。对待滤波像素块及其对应像素块滤波,得到待滤波像素块滤波后的像素。
实施例12
本发明实施例提供了一种全景图像像素块滤波方法。
本实施例提供的全景图像像素块滤波方法包括:
根据紧邻全景图像球面分割线一侧的待滤波像素块所在图像对应的投影格式信息,对所述待滤波像素块中的标称点沿所述球面分割线方向的位置进行伸缩变换,确定所述标称点在紧邻所述球面分割线另一侧的对应标称点的位置;由所述对应标称点确定所述待滤波像素块在紧邻所述球面分割线另一侧的对应像素块;对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
本实施例中所述待滤波像素块所在的图像为3x2的六面体格式,如图20所示。图21给出了一个待滤波像素块B及其在全景图像球面分割线另一侧的对应像素块B’。假设图像在X-Y坐标系下采样率相同,由所述格式信息可以得到图像在图21中的X’-Y’坐标系下采样率不同,且有以下关系:
Figure PCTCN2019083033-appb-000001
Figure PCTCN2019083033-appb-000002
β=tan -1x
其中x∈(-1,1),y∈(-1,1),x′∈(-1,1),y′∈(-1,1)。
如图22所示所述标称点为待滤波像素块B中最靠近面边界一行像素的几何中点b,根据该点横坐标计算出像素块B’中对应标称点b’的纵坐标。计算方法如下:
x=f -1(x′)
y′=g(x)
其中f -1(x′)表示f(x)的反函数。
图23给出了对应像素块B’中的对应标称点b’,对B’中的对应标称点b’的纵坐标取有限精度得到分像素精度,例如1/2像素精度、1/4像素精度、1/8像素精度、1/16像素精度等。其余非对应标称点所在位置的像素以b’为基准按照整像素间隔获取,非整像素位置的像素由周围整像素位置的像素插值得到。对待滤波像素块及其对应像素块滤波,得到待滤波像素块滤波后的像素。
实施例13
本发明实施例提供了一种全景图像像素块滤波方法。
本实施例提供的全景图像像素块滤波方法包括:
根据紧邻全景图像球面分割线一侧的待滤波像素块所在图像对应的投影格式信息,对所述待滤波像素块中的标称点沿所述球面分割线方向的位置进行伸缩变换,确定所述标称点在紧邻所述球面分割线另一侧的对应标称点的位置;由所述对应标称点确定所述待滤波像素块在紧邻所述球面分割线另一侧的对应像素块;对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
本实施例中所述待滤波像素块所在的图像为3x2的六面体格式,如图20所示。图21给出了一个待滤波像素块B及其在全景图像球面分割线另一侧的对应像素块B’。假设图像在X-Y坐标系下采样率相同,由所述格式信息可以得到图像在图21中的X’-Y’坐标系下采样率不同,且有以下关系:
Figure PCTCN2019083033-appb-000003
Figure PCTCN2019083033-appb-000004
其中x∈(-1,1),y∈(-1,1),x′∈(-1,1),y′∈(-1,1)。
如图22所示所述标称点为待滤波像素块B中最靠近面边界一行像素的几何中点b,根据该点横坐标计算出像素块B’中对应标称点b’的纵坐标。计算方法如下:
x=f -1(x′)
y′=g(x)
其中f -1(x′)表示f(x)的反函数。
图23给出了对应像素块B’中的对应标称点b’,对B’中的对应标称点b’的纵坐标取有限精度得到整像素精度。其余非对应标称点所在位置的像素以b’为基准按照整像素间隔获取。对待滤波像素块及其对应像素块滤波,得到待滤波像素块滤波后的像素。
实施例14
本发明实施例提供了一种全景图像像素块滤波方法。
本实施例提供的全景图像像素块滤波方法包括:
根据紧邻全景图像球面分割线一侧的待滤波像素块所在图像对应的投影格式信息,对所述待滤波像素块中的标称点沿所述球面分割线方向的位置进行伸缩变换,确定所述标称点在紧邻所述球面分割线另一侧的对应标称点的位置;由所述对应标称点确定所述待滤波像素块在紧邻所述球面分割线另一侧的对应像素块;对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
本实施例中所述待滤波像素块所在的图像为3x2的六面体格式,如图20所示。图21给出了一个待滤波像素块B及其在全景图像球面分割线另一侧的对应像素块B’。假设图像在X-Y坐标系下采样率相同,由所述格式信息可以得到图像在图21中的X’-Y’坐标系下采样率不同,且有以下关系:
Figure PCTCN2019083033-appb-000005
Figure PCTCN2019083033-appb-000006
其中x∈(-1,1),y∈(-1,1),x′∈(-1,1),y′∈(-1,1)。
如图22所示所述标称点为待滤波像素块B中最靠近面边界一行像素的几何中点b,根据该点横坐标计算出像素块B’中对应标称点b’的纵坐标。计算方法如下:
x=f -1(x′)
y′=g(x)
其中f -1(x′)表示f(x)的反函数。
图23给出了对应像素块B’中的对应标称点b’,对B’中的对应标称点b’的纵坐标取有限精度得到分像素精度,例如1/2像素精度、1/4像素精度、1/8像素精度、1/16像素精度等。其余非对应标称点所在位置的像素以b’为基准按照整像素间隔获取,非整像素位置的像素由周围整像素位置的像素插值得到。对待滤波像素块及其对应像素块滤波,得到待滤波像素块滤波后的像素。
实施例15
本发明实施例提供了一种全景图像像素块滤波方法。
本实施例提供的全景图像像素块滤波方法包括:
对紧邻全景图像球面分割线一侧的待滤波像素块,根据所述待滤波像素块所在图像对应的投影格式信息,确定所述待滤波像素块在紧邻所述球面分割线另一侧的对应像素块的位置;根据所述待滤波像素块及其对应像素块的编码信息确定滤波方法;其中所述待滤波像素块及其对应像素块中的编码信息中的运动信息始终认为相同;对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
图27给出了一个待滤波像素块B及其在全景图像上的对应像素块B’。在图27中像素块B的横坐标和像素块B’的横坐标是对齐的,因此不需要对像素块B的横坐标进行伸缩变换。根据所述待滤波像素块及其对应像素块的编码信息确定滤波方法,其中待滤波像素块及其对应像素块的编码信息中的运动信息不需要获取,认为待滤波像素块及其对应像素块的运动信息始终相等。对所述待滤波像素块及其对应像素块滤波,得到待滤波像素块滤波后的像素。
实施例16
本发明实施例提供了一种全景图像像素块滤波方法。
本实施例提供的全景图像像素块滤波方法包括:
对紧邻全景图像球面分割线一侧的待滤波像素块,根据所述待滤波像素块所在图像对应的投影格式信息,确定所述待滤波像素块在紧邻所述球面分割线另一侧的对应像素块的位置;根据所述待滤波像素块及其对应像素块的编码信息确定滤波方法;其中所述待滤波像素块及其对应像素块中的编码信息中的运动信息始终认为相同;对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
图28给出了一个经纬图格式上的待滤波像素块B及其对应像素块B’。在图28中像素块B的纵坐标和像素块B’的纵坐标是对齐的,因此不需要对像素块B的横坐标进行伸缩变换。根据所述待滤波像素块及其对应像素块的编码信息确定滤波方法,认为待滤波像素块及其对应像素块的编码信息中的运动信息始终相等。对所述待滤波像素块及其对应像素块滤波,得到待滤波像素块滤波后的像素。
实施例17
本发明实施例提供了一种全景图像像素块滤波方法。
本实施例提供的全景图像像素块滤波方法包括:
根据紧邻全景图像球面分割线一侧的待滤波像素块所在图像对应的投影格式信息,对所述待滤波像素块中的标称点沿所述球面分割线方向的位置进行伸缩变换,确定所述标称点在紧邻所述球面分割线另一侧的对应标称点的位置;由所述对应标称点确定所述待滤波像素块在紧邻所述球面分割线另一侧的对应像素块;对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
本实施例中所述滤波指去块滤波。在对待滤波像素块及其对应像素块滤波前,需要确定对应的滤波方法,在确定滤波方法时,将待滤波像素块及其对应像素块的边界强度设为1,确定滤波方法不需要获取待滤波像素块及其对应像素块的编码信息,而是使用待滤波像素块所在图像的量化步长作为待滤波像素块的量化步长,对应像素块所在图像的量化步长作为对应像素块的量化步长。根据边界强度、量化步长以及待滤波像素块及其对应像素块中的像素特性确定编码方法,对待滤波像素块及其对应像素块 滤波,得到待滤波像素块滤波后的像素。
实施例18
本发明实施例提供了一种全景图像像素块滤波方法。
本实施例提供的全景图像像素块滤波方法包括:
根据紧邻全景图像球面分割线一侧的待滤波像素块所在图像对应的投影格式信息,对所述待滤波像素块中的标称点沿所述球面分割线方向的位置进行伸缩变换,确定所述标称点在紧邻所述球面分割线另一侧的对应标称点的位置;由所述对应标称点确定所述待滤波像素块在紧邻所述球面分割线另一侧的对应像素块;对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
在对待滤波像素块及其对应像素块滤波前,需要确定对应的滤波方法,在确定滤波方法时,将待滤波像素块及其对应像素块的边界强度设为2,确定滤波方法不需要获取待滤波像素块及其对应像素块的编码信息。使用待滤波像素块所在图像的量化步长作为待滤波像素块的量化步长,对应像素块所在图像的量化步长作为对应像素块的量化步长。根据边界强度、量化步长以及待滤波像素块及其对应像素块中的像素特性确定编码方法,对待滤波像素块及其对应像素块滤波,得到待滤波像素块滤波后的像素。
实施例19
本发明实施例提供了一种全景图像像素块滤波方法。
本实施例提供的全景图像像素块滤波方法包括:
根据紧邻全景图像球面分割线一侧的待滤波像素块所在图像对应的投影格式信息,对所述待滤波像素块中的标称点沿所述球面分割线方向的位置进行伸缩变换,确定所述标称点在紧邻所述球面分割线另一侧的对应标称点的位置;由所述对应标称点确定所述待滤波像素块在紧邻所述球面分割线另一侧的对应像素块;对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
在对待滤波像素块及其对应像素块滤波前,需要确定对应的滤波方法,在确定滤波方法时,待滤波像素块及其对应像素块的边界强度根据两像素块的编码信息决定,若两个像素块中至少有一个块的编码信息中的预测模式为帧内预测,则边界强度为2,否则边界强度为1。使用待滤波像素块所在图像的量化步长作为待滤波像素块的量化步长,对应像素块所在图像的量化步长作为对应像素块的量化步长。根据边界强度、量化步长以及待滤波像素块及其对应像素块中的像素特性确定编码方法,对待滤波像素块及其对应像素块滤波,得到待滤波像素块滤波后的像素。
实施例20
本发明实施例提供了一种全景图像像素块滤波方法。
本实施例提供的全景图像像素块滤波方法包括:
根据紧邻全景图像球面分割线一侧的待滤波像素块所在图像对应的投影格式信息,对所述待滤波像素块中的标称点沿所述球面分割线方向的位置进行伸缩变换,确定所述标称点在紧邻所述球面分割线另一侧的对应标称点的位置;由所述对应标称点确定所述待滤波像素块在紧邻所述球面分割线另一侧的对应像素块;对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
在对待滤波像素块及其对应像素块滤波前,需要确定对应的滤波方法,在确定滤 波方法时,待滤波像素块及其对应像素块的边界强度根据两像素块的编码信息决定,若两个像素块中至少有一个块的编码信息中的预测模式为帧内预测,则边界强度为2,否则边界强度为1。确定滤波像素所需的量化步长从待滤波像素块及其对应像素块的编码信息中取出。根据边界强度、量化步长以及待滤波像素块及其对应像素块中的像素特性确定编码方法,对待滤波像素块及其对应像素块滤波,得到待滤波像素块滤波后的像素。
实施例21
本发明实施例提供了一种全景图像像素块滤波方法。
本实施例提供的全景图像像素块滤波方法包括:
根据紧邻全景图像球面分割线一侧的待滤波像素块所在图像对应的投影格式信息,对所述待滤波像素块中的标称点沿所述球面分割线方向的位置进行伸缩变换,确定所述标称点在紧邻所述球面分割线另一侧的对应标称点的位置;由所述对应标称点确定所述待滤波像素块在紧邻所述球面分割线另一侧的对应像素块;对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
在对待滤波像素块及其对应像素块滤波前,需要确定对应的滤波方法,在确定滤波方法时,待滤波像素块及其对应像素块的边界强度根据两像素块的编码信息决定,若两个像素块中至少有一个块的编码信息中的预测模式为帧内预测,则边界强度为2,否则边界强度为1。确定滤波方法所需的量化步长不是从待滤波像素块及其对应像素块中直接取出,而是使用待滤波像素块所在编码区域的量化步长和待滤波像素块的量化步长的较大值作为待滤波像素块的量化步长;对应像素块所在编码区域的量化步长和对应像素块的量化步长作为对应像素块的量化步长。根据边界强度、量化步长以及待滤波像素块及其对应像素块中的像素特性确定编码方法,对待滤波像素块及其对应像素块滤波,得到待滤波像素块滤波后的像素。
实施例22
本发明实施例提供了一种全景图像像素块滤波装置。
本实施例提供的全景图像像素块滤波装置,包括以下模块:
对应标称点位置导出模块:该模块的输入为紧邻全景图像球面分割线一侧的待滤波像素块中标称点的位置信息以及所述待滤波像素块所在图像对应的投影格式信息;输出为所述标称点在全景图像中对应标称点的位置;根据紧邻全景图像球面分割线一侧的待滤波像素块所在图像对应的投影格式信息,对所述待滤波像素块中的标称点沿所述球面分割线方向的位置进行伸缩变换,确定所述标称点在紧邻所述球面分割线另一侧的对应标称点的位置;
图29给出了本实施例提供的全景图像像素块滤波装置示意图。在该模块中对对应标称点位置取有限精度得到整像素位置。
对应像素块位置导出模块:该模块输入为对应标称点的位置,输出为待滤波像素块在全景图像中对应像素块的位置;该模块用于根据对应标称点的位置确定对应像素块中其余非对应标称点的位置;
该模块中对非标称点位置的确定方法为:以对应像素块中对应标称点为基准,按整像素间隔确定所述非对应标称点的位置。
像素滤波模块:该模块输入为待滤波像素块位置及其对应像素块的位置,输出为 待滤波像素块滤波后的像素,该模块用于对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
该模块需要获取待滤波像素块及其对应像素块的编码信息,根据待滤波像素块及其对应像素块的编码信息和像素确定滤波方法,对待滤波像素块及其对应像素块滤波,得到待滤波像素块滤波后的结果。
实施例23
本发明实施例提供了一种全景图像像素块滤波装置。
本实施例提供的全景图像像素块滤波装置,包括以下模块:
对应像素块位置导出模块:该模块输入为紧邻全景图像球面分割线一侧的待滤波像素块的位置信息以及所述待滤波像素块所在图像对应的投影格式信息,输出为待滤波像素块的对应像素块位置;
滤波方法确定模块,用于确定滤波方法,输入为待滤波像素块及其对应像素块的位置信息,根据所述待滤波像素块及其对应像素块的编码信息确定滤波方法,确定滤波方法为以下之一种:认为所述待滤波像素块及其对应像素块的编码信息中的运动信息相同,或者不使用所述待滤波像素块及其对应像素块的编码信息中的运动信息,输出为该模块确定的滤波方法;
像素滤波模块,用于对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
图30给出了本实施例提供的全景图像像素块滤波装置示意图。所述对应像素块位置导出模块在获得对应像素块位置后,将该位置传递给滤波方法确定模块,滤波方法确定模块确定滤波方法并将确定的滤波方法传递给待滤波像素块,像素滤波模块根据输入的滤波方法对输入的像素块滤波,获得待滤波像素块滤波后的结果。
实施例24
本发明实施例提供了一种全景图像像素块滤波装置。
本实施例提供的全景图像像素块滤波装置,包括以下模块:
对应标称点位置导出模块:该模块的输入为紧邻全景图像球面分割线一侧的待滤波像素块中标称点的位置信息以及所述待滤波像素块所在图像对应的投影格式信息;输出为所述标称点在全景图像中对应标称点的位置;根据紧邻全景图像球面分割线一侧的待滤波像素块所在图像对应的投影格式信息,对所述待滤波像素块中的标称点沿所述球面分割线方向的位置进行伸缩变换,确定所述标称点在紧邻所述球面分割线另一侧的对应标称点的位置;
对应像素块位置导出模块:该模块输入为对应标称点的位置,输出为待滤波像素块在全景图像中对应像素块的位置;该模块用于根据对应标称点的位置确定对应像素块中其余非对应标称点的位置;
像素滤波模块:该模块输入为待滤波像素块位置及其对应像素块的位置,输出为待滤波像素块滤波后的像素,该模块用于对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
如图31所示,所述对应标称点位置导出模块中包含对应标称点位置取整模块,该模块对对应标称点坐标取有限精度得到分像素精度,例如1/2像素精度、1/4像素精度、1/8像素精度、1/16像素精度等。
对应像素块位置导出模块中还包括非对应标称点位置导出模块,如图32所示。非对应标称点位置导出模块对非标称点位置的确定方法为:以对应像素块中对应标称点为基准,按整像素间隔确定所述非对应标称点的位置。
像素滤波模块中还包括编码信息获取模块以及滤波方法确定模块,如图33所示。编码信息获取模块根据输入的待滤波像素块及其对应像素块的位置获取所述待滤波像素块及其对应像素块的编码信息;滤波方法确定模块根据待滤波像素块及其对应像素块的编码信息和像素特性确定滤波方法。根据确定好的滤波方法对待滤波像素块及其对应块滤波,得到待滤波像素滤波后的结果。
实施例25
本发明实施例提供了一种全景图像像素块滤波装置。
本实施例提供的全景图像像素块滤波装置,包括以下模块:
对应标称点位置导出模块:该模块的输入为紧邻全景图像球面分割线一侧的待滤波像素块中标称点的位置信息以及所述待滤波像素块所在图像对应的投影格式信息;输出为所述标称点在全景图像中对应标称点的位置;根据紧邻全景图像球面分割线一侧的待滤波像素块所在图像对应的投影格式信息,对所述待滤波像素块中的标称点沿所述球面分割线方向的位置进行伸缩变换,确定所述标称点在紧邻所述球面分割线另一侧的对应标称点的位置;
对应像素块位置导出模块:该模块输入为对应标称点的位置,输出为待滤波像素块在全景图像中对应像素块的位置;该模块用于根据对应标称点的位置确定对应像素块中其余非对应标称点的位置;
像素滤波模块:该模块输入为待滤波像素块位置及其对应像素块的位置,输出为待滤波像素块滤波后的像素,该模块用于对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
在像素滤波模块中,对待滤波像素块及其对应像素块滤波前,需要确定对应的滤波方法,在确定滤波方法时,将待滤波像素块及其对应像素块的边界强度设为2,确定滤波方法不需要获取待滤波像素块及其对应像素块的编码信息。使用待滤波像素块所在图像的量化步长作为待滤波像素块的量化步长,对应像素块所在图像的量化步长作为对应像素块的量化步长。根据边界强度、量化步长以及待滤波像素块及其对应像素块中的像素特性确定编码方法,对待滤波像素块及其对应像素块滤波,得到待滤波像素块滤波后的像素。
实施例26
本发明实施例提供了一种全景图像像素块滤波装置。
本实施例提供的全景图像像素块滤波装置,包括以下模块:
对应标称点位置导出模块:该模块的输入为紧邻全景图像球面分割线一侧的待滤波像素块中标称点的位置信息以及所述待滤波像素块所在图像对应的投影格式信息;输出为所述标称点在全景图像中对应标称点的位置;根据紧邻全景图像球面分割线一侧的待滤波像素块所在图像对应的投影格式信息,对所述待滤波像素块中的标称点沿所述球面分割线方向的位置进行伸缩变换,确定所述标称点在紧邻所述球面分割线另一侧的对应标称点的位置;
对应像素块位置导出模块:该模块输入为对应标称点的位置,输出为待滤波像素 块在全景图像中对应像素块的位置;该模块用于根据对应标称点的位置确定对应像素块中其余非对应标称点的位置;
像素滤波模块:该模块输入为待滤波像素块位置及其对应像素块的位置,输出为待滤波像素块滤波后的像素,该模块用于对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
在像素滤波模块中,对待滤波像素块及其对应像素块滤波前,需要确定对应的滤波方法,在确定滤波方法时,将待滤波像素块及其对应像素块的边界强度设为1,确定滤波方法不需要获取待滤波像素块及其对应像素块的编码信息。使用待滤波像素块所在图像的量化步长作为待滤波像素块的量化步长,对应像素块所在图像的量化步长作为对应像素块的量化步长。根据边界强度、量化步长以及待滤波像素块及其对应像素块中的像素特性确定编码方法,对待滤波像素块及其对应像素块滤波,得到待滤波像素块滤波后的像素。
实施例27
本发明实施例提供了一种全景图像像素块滤波装置。
本实施例提供的全景图像像素块滤波装置,包括以下模块:
对应像素块位置导出模块:该模块输入为紧邻全景图像球面分割线一侧的待滤波像素块的位置信息以及所述待滤波像素块所在图像对应的投影格式信息,输出为待滤波像素块的对应像素块位置;
编码信息获取模块,用于获取所述待滤波像素块及其对应像素块的编码信息,其中所述待滤波像素块及其对应像素块的编码信息中的运动信息始终认为相同;
像素滤波模块,用于对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
在编码信息获取模块中,待滤波像素块及其对应像素块的编码信息中的运动信息不需要获取,认为待滤波像素块及其对应像素块的运动信息始终相等。
实施例28
本发明实施例提供了一种全景图像像素块滤波装置。
本实施例提供的全景图像像素块滤波装置,包括以下模块:
对应标称点位置导出模块:该模块的输入为紧邻全景图像球面分割线一侧的待滤波像素块中标称点的位置信息以及所述待滤波像素块所在图像对应的投影格式信息;输出为所述标称点在全景图像中对应标称点的位置;根据紧邻全景图像球面分割线一侧的待滤波像素块所在图像对应的投影格式信息,对所述待滤波像素块中的标称点沿所述球面分割线方向的位置进行伸缩变换,确定所述标称点在紧邻所述球面分割线另一侧的对应标称点的位置;
对应像素块位置导出模块:该模块输入为对应标称点的位置,输出为待滤波像素块在全景图像中对应像素块的位置;该模块用于根据对应标称点的位置确定对应像素块中其余非对应标称点的位置;
像素滤波模块:该模块输入为待滤波像素块位置及其对应像素块的位置,输出为待滤波像素块滤波后的像素,该模块用于对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
在对应标称点位置导出模块中,对对应标称点的坐标取整获得整像素位置;在对 应像素块位置导出模块中,对应像素块中的其余非标称点的位置以对应标称点为基准,按整像素间隔获取。
在像素滤波模块中,对待滤波像素块及其对应像素块滤波前,需要确定对应的滤波方法,在确定滤波方法时,将待滤波像素块及其对应像素块的边界强度设为1,确定滤波方法不需要获取待滤波像素块及其对应像素块的编码信息。使用待滤波像素块所在图像的量化步长作为待滤波像素块的量化步长,对应像素块所在图像的量化步长作为对应像素块的量化步长。根据边界强度、量化步长以及待滤波像素块及其对应像素块中的像素特性确定编码方法,对待滤波像素块及其对应像素块滤波,得到待滤波像素块滤波后的像素。
图34给出了未滤波前全景图像的一个观看区域中的伪像现象,通过使用本实施例的全景图像像素块滤波装置对全景图像待滤波像素块滤波后生成的观看区域如图35所示,可以看出编码造成的伪像大大减弱。
实施例29
本发明实施例提供了一种全景图像像素块滤波方法。
本实施例提供的全景图像像素块滤波方法包括:
对紧邻全景图像球面分割线一侧的待滤波像素块,根据所述待滤波像素块所在图像对应的投影格式信息,确定所述待滤波像素块在紧邻所述球面分割线另一侧的对应像素块的位置;根据所述待滤波像素块及其对应像素块的编码信息确定滤波方法,确定滤波方法为以下之一种:认为所述待滤波像素块及其对应像素块的编码信息中的运动信息相同,或者不使用所述待滤波像素块及其对应像素块的编码信息中的运动信息;对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
在本实施例中,对上述根据所述待滤波像素块及其对应像素块的编码信息确定滤波方法这一过程进行了更加详细的描述。在确定滤波方法时,需要确定待滤波像素块及其对应像素块的边界强度,在确定边界强度时不使用编码信息中的运动信息或者认为编码信息中的运动信息相等,若两个像素块中至少有一个块的编码信息中的预测模式为帧内预测,则边界强度为2,否则边界强度为1。使用编码信息中的QP来确定滤波方法的判决门限。根据边界强度、判决门限以及待滤波像素块及其对应像素块中的像素特性确定编码方法,对待滤波像素块及其对应像素块滤波,得到待滤波像素块滤波后的像素。
实施例30
本发明实施例提供了一种全景图像像素块滤波方法。
本实施例提供的全景图像像素块滤波方法包括:
对紧邻全景图像球面分割线一侧的待滤波像素块,根据所述待滤波像素块所在图像对应的投影格式信息,确定所述待滤波像素块在紧邻所述球面分割线另一侧的对应像素块的位置;根据所述待滤波像素块及其对应像素块的编码信息确定滤波方法,确定滤波方法为以下之一种:认为所述待滤波像素块及其对应像素块的编码信息中的运动信息相同,或者不使用所述待滤波像素块及其对应像素块的编码信息中的运动信息;对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
在本实施例中,对上述根据所述待滤波像素块及其对应像素块的编码信息确定滤波方法这一过程进行了更加详细的描述。在确定滤波方法时,需要确定待滤波像素块 及其对应像素块的边界强度,在确定边界强度时不使用编码信息中的运动信息或者认为编码信息中的运动信息相等,将边界强度设为固定值1或者设为固定值2,使用编码信息中的QP来确定滤波方法的判决门限。根据边界强度、判决门限以及待滤波像素块及其对应像素块中的像素特性确定编码方法,对待滤波像素块及其对应像素块滤波,得到待滤波像素块滤波后的像素。
实施例31
本发明实施例提供了一种全景图像像素块滤波方法。
本实施例提供的全景图像像素块滤波方法包括:
对紧邻全景图像球面分割线一侧的待滤波像素块,根据所述待滤波像素块所在图像对应的投影格式信息,确定所述待滤波像素块在紧邻所述球面分割线另一侧的对应像素块的位置;根据所述待滤波像素块及其对应像素块的编码信息确定滤波方法,确定滤波方法为以下之一种:认为所述待滤波像素块及其对应像素块的编码信息中的运动信息相同,或者不使用所述待滤波像素块及其对应像素块的编码信息中的运动信息;对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
在本实施例中,对上述根据所述待滤波像素块及其对应像素块的编码信息确定滤波方法这一过程进行了更加详细的描述。在确定滤波方法时,需要确定待滤波像素块及其对应像素块的边界强度,本实施例中在确定待滤波像素块及其对应像素块的边界强度时仅仅使用了帧内/帧间预测模式这一判决条件或者直接将边界强度设为固定大小的值,图36(a)展示了判断边界强度时使用帧内/帧间预测模式这一判决条件的流程图,若为帧内预测则边界强度为2,否则为1;图36(b)展示了边界强度判决的另一种方法,即边界强度固定为2或固定为1。确定边界强度之后还需要根据待滤波像素块及其对应块的编码信息中的QP来确定各滤波方法的判决门限值,针对不同的量化情况确定不同的门限值能够更好的提升滤波效果。
传统的滤波技术在判决边界强度时的流程较为复杂同时需要存储大量的编码信息,例如图37展示了HEVC中的去块滤波技术中边界强度的判决流程,在判断中需要的比特数分别为:帧内/帧间预测flag,1bit;非零变换系数flag,1bit;运动信息:运动矢量(MVx,MVy)其中MVx和MVy各需要16bit,参考帧的index:4bit,因此在双假设预测时共需要72bit的运动信息。而本实施例中仅仅使用了1bit(帧内/帧间预测flag)甚至不使用任何编码信息。
在现有的编解码技术中,编码信息是按照块来存储的,例如以4*4大小的块来存储编码信息。本发明相对于HEVC中的去块滤波技术对每个4*4块节省了73bit或74bit。每个4*4块中的像素数占用的大小为16*bitdepth bit,其中bitdepth表示每个像素使用的比特数,如8bit或10bit.当为8bit时一个4*4大小的像素块的像素使用的总比特数为128bit,加上编码信息共202bit,因此HEVC中的去块滤波技术对于每个4*4块使用了202bit,而本发明实施例中的方法仅使用128bit或129bit,和HEVC中的去块滤波技术相比节省了36%的行缓存。
最后应说明的是:以上实施例仅用以说明本发明的技术方案,而非对其限制;尽管参照前述实施例对本发明进行了详细的说明,本领域的普通技术人员应当理解:其依然可以对前述各实施例所记载的技术方案进行修改,或者对其中部分技术特征进行 等同替换;而这些修改或者替换,并不使相应技术方案的本质脱离本发明各实施例技术方案的范围。

Claims (18)

  1. 一种全景图像滤波方法,其特征在于,包括:
    根据全景投影展开图像的格式信息,确定在全景投影展开图像中存在的在全景球面上相邻、但在投影展开图像上不相邻的面边界区域对的位置;
    按照所述面边界区域对中一个特定的面边界区域在全景投影展开图像上的方向,确定面边界区域对中第一维和第二维滤波处理的方向,所述面边界区域对确定第一维和第二维滤波处理方向的方法与非面边界区域在全景投影展开图像上确定第一维和第二维滤波处理的方向一致;
    对所述面边界区域对依次进行第一维和第二维滤波处理。
  2. 根据权利要求1中的全景图像像素滤波方法,其特征还在于,所述特定的面边界区域为所述面边界区域对中在全景投影展开图像上较后编码或解码处理的面边界区域。
  3. 根据权利要求1中的全景图像像素滤波方法,其特征还在于,对于同属于两个面边界区域的共同子区域,所述共同子区域滤波后的像素由所述两个面边界区域中在全景展开投影图像上较后编码或解码处理的面边界区域所对应的子区域滤波后的像素来确定。
  4. 一种全景图像滤波装置,其特征在于,包括以下模块:
    待滤波面边界区域对位置导出模块:该模块的输入为全景投影展开图像的格式信息,根据格式信息确定在全景投影展开图像中存在的在全景球面上相邻、但在投影展开图像上不相邻的面边界区域对的位置,该模块的输出为所述待滤波面边界区域对在全景投影展开图像中的位置;
    滤波处理方向确定模块:该模块输入待滤波面边界区域对及其在全景投影展开图像中的位置,按照所述面边界区域对中一个特定的面边界区域在全景投影展开图像上的方向,确定面边界区域对中第一维和第二维滤波处理的方向,所述面边界区域对确定第一维和第二维滤波处理方向的方法与非面边界区域在全景投影展开图像上确定第一维和第二维滤波处理的方向一致;该模块输出待滤波面边界区域对的第一维和第二维滤波处理的方向;
    滤波处理模块:输入为待滤波面边界区域对的第一维和第二维滤波处理的方向,对所述面边界区域对依次进行第一维和第二维滤波处理。
  5. 根据权利要求4中的全景图像滤波装置,其特征还在于,所述滤波处理方向确定模块中的一个特定面边界区域为所述面边界区域对中在全景投影展开图像上较后编码或解码处理的面边界区域。
  6. 根据权利要求4中的全景图像滤波装置,其特征还在于,所述滤波处理模块中,对于同属于两个面边界区域的共同子区域,所述共同子区域滤波后的像素由所述两个面边界区域中在全景展开投影图像上较后编码/解码处理的面边界区域所对应的子区域滤波后的像素来确定。
  7. 一种全景图像像素块滤波方法,其特征在于,包括:
    根据紧邻全景图像球面分割线一侧的待滤波像素块所在图像对应的投影格式信息,对所述待滤波像素块中的标称点沿所述球面分割线方向的位置进行伸缩变换,确定所述标称点在紧邻所述球面分割线另一侧的对应标称点的位置;由所述对应标称点确定所述待滤波像素块在紧邻所述球面分割线另一侧的对应像素块;对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
  8. 根据权利7中的全景图像像素块滤波方法,其特征还在于,所述待滤波像素块的 对应像素块中对应标称点位置的确定方法为以下情况之一:
    1)对所述对应像素块中对应标称点的位置取有限精度得到整像素位置;
    2)对所述对应像素块中对应标称点的位置取有限精度得到分像素位置。
  9. 根据权利7中的全景图像像素块滤波方法,其特征还在于,所述待滤波像素块的对应像素块中其余非对应标称点的位置的确定方法为:以所述对应像素块中对应标称点为基准,按整像素间隔确定所述对应像素块中非对应标称点的位置。
  10. 根据权利7中的全景图像像素块滤波方法,其特征还在于,包括:
    获取所述待滤波像素块及其对应像素块的编码信息,根据所述待滤波像素块及其对应像素块的编码信息确定滤波方法。
  11. 根据权利10中的全景图像像素块滤波方法,其特征还在于,包括:
    所述待滤波像素块对应像素块的编码信息的获取方法为以下情况之一:
    1)所述编码信息来自所述对应像素块中对应标称点所在的编码单元;
    2)所述编码信息来自包含所述对应像素块中像素数目最多的编码单元。
  12. 一种全景图像像素块滤波方法,其特征在于,包括:
    对紧邻全景图像球面分割线一侧的待滤波像素块,根据所述待滤波像素块所在图像对应的投影格式信息,确定所述待滤波像素块在紧邻所述球面分割线另一侧的对应像素块的位置;根据所述待滤波像素块及其对应像素块的编码信息确定滤波方法,确定滤波方法为以下之一种:认为所述待滤波像素块及其对应像素块的编码信息中的运动信息相同,或者不使用所述待滤波像素块及其对应像素块的编码信息中的运动信息;对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
  13. 一种全景图像像素块滤波装置,其特征在于,包括以下模块:
    对应标称点位置导出模块:该模块的输入为紧邻全景图像球面分割线一侧的待滤波像素块中标称点的位置信息以及所述待滤波像素块所在图像对应的投影格式信息;输出为所述标称点在全景图像中对应标称点的位置;根据紧邻全景图像球面分割线一侧的待滤波像素块所在图像对应的投影格式信息,对所述待滤波像素块中的标称点沿所述球面分割线方向的位置进行伸缩变换,确定所述标称点在紧邻所述球面分割线另一侧的对应标称点的位置;
    对应像素块位置导出模块:该模块输入为对应标称点的位置,输出为待滤波像素块在全景图像中对应像素块的位置;该模块用于根据对应标称点的位置确定对应像素块中其余非对应标称点的位置;
    像素滤波模块:该模块输入为待滤波像素块位置及其对应像素块的位置,输出为待滤波像素块滤波后的像素,该模块用于对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
  14. 根据权利要求13中的全景图像像素块滤波装置,其特征还在于,所述对应标称点位置导出模块中还包括对应标称点位置取整模块,所述对应标称点位置取整模块中对应标称点位置的取整方法为以下情况之一:
    1)对所述对应像素块中对应标称点的位置取有限精度得到整像素位置;
    2)对所述对应像素块中对应标称点的位置取有限精度得到分像素位置。
  15. 根据权利要求13中的全景图像像素块滤波装置,其特征还在于,所述对应像素块位置导出模块中还包括非对应标称点位置导出模块,所述非对应标称点位置导出模块中非标称点位置的确定方法为:以对应像素块中对应标称点为基准,按整像素间隔确定 所述非对应标称点的位置。
  16. 根据权利要求13中的全景图像像素块滤波装置,其特征还在于,所述像素滤波模块中还包括以下子模块:
    编码信息获取模块,用于获取所述待滤波像素块及其对应像素块的编码信息;
    滤波方法确定模块,用于根据所述待滤波像素块及其对应像素块的编码信息和像素特性确定滤波方法。
  17. 根据权利要求16中的编码信息获取模块,其特征还在于,所述待滤波像素块对应像素块的编码信息的获取方法为以下情况之一:
    1)所述编码信息来自所述对应像素块中对应标称点所在的编码单元;
    2)所述编码信息来自包含所述对应像素块中像素数目最多的编码单元。
  18. 一种全景图像像素块滤波装置,其特征在于,包括以下模块:
    对应像素块位置导出模块:该模块输入为紧邻全景图像球面分割线一侧的待滤波像素块的位置信息以及所述待滤波像素块所在图像对应的投影格式信息,输出为待滤波像素块的对应像素块位置;
    滤波方法确定模块,用于确定滤波方法,输入为待滤波像素块及其对应像素块的位置信息,根据所述待滤波像素块及其对应像素块的编码信息确定滤波方法,确定滤波方法为以下之一种:认为所述待滤波像素块及其对应像素块的编码信息中的运动信息相同,或者不使用所述待滤波像素块及其对应像素块的编码信息中的运动信息,输出为该模块确定的滤波方法;
    像素滤波模块,用于对所述待滤波像素块及其对应像素块滤波,得到所述待滤波像素块滤波后的结果。
PCT/CN2019/083033 2018-07-02 2019-04-17 一种全景图像滤波方法及装置 WO2020007094A1 (zh)

Applications Claiming Priority (4)

Application Number Priority Date Filing Date Title
CN201810709252.X 2018-07-02
CN201810709252 2018-07-02
CN201811111831 2018-09-24
CN201811111831.0 2018-09-24

Publications (1)

Publication Number Publication Date
WO2020007094A1 true WO2020007094A1 (zh) 2020-01-09

Family

ID=69060772

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2019/083033 WO2020007094A1 (zh) 2018-07-02 2019-04-17 一种全景图像滤波方法及装置

Country Status (1)

Country Link
WO (1) WO2020007094A1 (zh)

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101079151A (zh) * 2006-10-13 2007-11-28 浙江师范大学 一种基于序列静态图像的360°环视全景生成方法
CN107948547A (zh) * 2017-12-29 2018-04-20 北京奇艺世纪科技有限公司 全景视频拼接的处理方法、装置及电子设备
CN108012153A (zh) * 2016-10-17 2018-05-08 联发科技股份有限公司 一种编解码方法及装置
CN108111851A (zh) * 2016-11-25 2018-06-01 华为技术有限公司 一种去块滤波方法及终端

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101079151A (zh) * 2006-10-13 2007-11-28 浙江师范大学 一种基于序列静态图像的360°环视全景生成方法
CN108012153A (zh) * 2016-10-17 2018-05-08 联发科技股份有限公司 一种编解码方法及装置
CN108111851A (zh) * 2016-11-25 2018-06-01 华为技术有限公司 一种去块滤波方法及终端
CN107948547A (zh) * 2017-12-29 2018-04-20 北京奇艺世纪科技有限公司 全景视频拼接的处理方法、装置及电子设备

Similar Documents

Publication Publication Date Title
CN107710761B (zh) 图像预测方法和相关设备
JP2023053247A (ja) 省略符号化を用いた映像符号化及び復号化装置及びその方法
CN114208197A (zh) 用于视频编解码的块划分方法
US11375199B2 (en) Interpolation filter for an inter prediction apparatus and method for video coding
CN110892719A (zh) 影像编码/解码方法及装置
CN108605144B (zh) 一种去除块效应的滤波方法及装置
CN109688413B (zh) 以支持辅助帧的视频编码格式编码视频流的方法和编码器
US20230063062A1 (en) Hardware codec accelerators for high-performance video encoding
US20240297969A1 (en) Image data encoding/decoding method and apparatus
CN109804631B (zh) 一种用于编码和解码视频信号的装置和方法
CN110971897B (zh) 色度分量的帧内预测模式的编码、解码方法、设备和系统
JP2024084840A (ja) 画像データ符号化/復号化方法及び装置
CN116760965B (zh) 全景视频编码方法、装置、计算机设备和存储介质
Lin et al. Residual in residual based convolutional neural network in-loop filter for AVS3
WO2020007094A1 (zh) 一种全景图像滤波方法及装置
CN110675401B (zh) 一种全景图像像素块滤波方法及装置
WO2019192377A1 (zh) 一种全景图像处理方法及装置
US11202082B2 (en) Image processing apparatus and method
CN110942428A (zh) 一种全景图像滤波方法及装置
CN111758258A (zh) 用于视频编码的帧间预测装置和方法
US20240267529A1 (en) Context-aware quantization for high-performance video encoding
Zhang et al. Recurrent Information Synthesis for Down-Sampling Based Video Coding
TW202034699A (zh) 圖像編碼裝置、圖像解碼裝置、圖像編碼方法、圖像解碼方法及程式

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: 19831092

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: 19831092

Country of ref document: EP

Kind code of ref document: A1