A method for correcting spatial consistency in multi-view surround shooting

Abstract

The present invention relates to the field of image processing technology, and in particular to a method for correcting spatial consistency of multi-view surround shooting, comprising: step S1, extracting frames from images of N camera positions to obtain a sequence frame file, and calculating an initial correction file and initial correction parameters of each camera position; step S2, re-extracting frames from the images of the N camera positions to obtain a new sequence frame file, and obtaining initial correction parameters of an abnormal camera position from the initial correction file; step S3, performing correction processing on a foreground image and a background image until the foreground image and the background image completely overlap, and obtaining new correction parameters based on the correction processing process; step S4, replacing the initial correction parameters of the abnormal camera position with the new correction parameters and correcting the image of the abnormal camera position using the new correction parameters. The present invention does not require erecting a correction rod, and obtains new correction parameters through image comparison and algorithm verification, thereby meeting video rendering requirements.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a correction method for multi-view surround shooting. Background Art

[0002] In the multi-view surround shooting process, the spatial consistency processing of multiple camera positions is a core technology. If the camera shooting angles are not corrected consistently, jitter will occur in the final video effect. The existing technology needs to place correction rods in the setup stage before shooting, extract still images of each camera position, and calculate and obtain the position and posture deviation of each camera position relative to the reference camera position based on the position of the correction rod in each camera position, so that each camera position can be corrected and processed in real time in subsequent shooting.

[0003] However, as filming progresses, the correction file parameters acquired initially change over time due to factors such as vibration of the camera's pan / tilt base, human contact, and unstable counterweights. This makes the correction file parameters no longer suitable for the current spatial consistency requirements. Due to live broadcasting and other reasons, the on-site environment often does not allow for the re-erecting of the correction pole and the re-creation of the correction file. This results in jitter in the rendered video. Therefore, the abnormal camera position must be discarded, sacrificing image quality to ensure the normal live broadcast. Summary of the Invention

[0004] The purpose of the present invention is to provide a method for correcting spatial consistency of multi-view surround shooting to solve the above technical problems;

[0005] A method for correcting spatial consistency of multi-view surround shooting, comprising:

[0006] Step S1, extracting frames from N camera positions to obtain a sequence frame file, and correcting the N camera positions in the sequence frame file based on a reference camera position to obtain an initial correction file and initial correction parameters for each camera position;

[0007] Step S2, re-extracting frames from the N camera positions to obtain a new sequence frame file, determining an abnormal camera position where a new offset image appears based on a playback result of the new sequence frame file, and obtaining initial correction parameters for the abnormal camera position from the initial correction file;

[0008] Step S3, using the image of the abnormal camera position in the sequence frame file as the background image, and using the image of the abnormal camera position in the new sequence frame file as the foreground image, performing correction processing on the foreground image and the background image until the foreground image and the background image completely overlap, and obtaining new correction parameters based on the correction processing process;

[0009] Step S4: replacing the initial correction parameters of the abnormal camera position with the new correction parameters and correcting the image of the abnormal camera position using the new correction parameters.

[0010] Preferably, the step S3 includes:

[0011] Step S31, establishing a virtual machine position, and importing the background image and the foreground image into the virtual machine position;

[0012] Step S32 , marking a first marking point A ( x1 , y1 ) and a second marking point B ( x2 , y2 ) on the background image, and marking a third marking point C ( x3 , y3 ) and a fourth marking point D ( x4 , y4 ) on the foreground image;

[0013] In step S33, the foreground image is corrected by scaling, rotating, and translating the foreground image through an affine transformation matrix so that the first marker point A (x1, y1) and the third marker point C (x3, y3) coincide with each other, and the second marker point B (x2, y2) and the fourth marker point D (x4, y4) coincide with each other.

[0014] Preferably, the first marking point A (x1, y1) and the third marking point C (x3, y3) are located at the same position of the background picture and the foreground picture; the second marking point B (x2, y2) and the fourth marking point D (x4, y4) are located at the same position of the background picture and the foreground picture.

[0015] Preferably, the scaling parameters, rotation parameters, and translation parameters calculated during the correction operation in step S33 are used as the new correction parameters.

[0016] Preferably, the calculation formula of the scaling parameter is:

[0017]

[0018] Wherein, s is the scaling parameter, s0 is the original scaling parameter in the original correction parameter, Δs is the scaling parameter increment, a3 is the length of the line segment between the third marking point C (x3, y3) and the fourth marking point D (x4, y4) after correction by the scaling parameter, and a1 is the length of the line segment between the first marking point A (x1, y1) and the second marking point B (x2, y2).

[0019] Preferably, the scaling parameter increment is increased multiple times when scaling the foreground image, and after each increase in the scaling parameter increment, it is determined whether a3 is equal to a1. If they are equal, the current scaling ratio is used as the scaling parameter. Otherwise, the scaling parameter increment is continued to be increased until a3 and a1 are equal. The calculation formula for the scaling parameter increment is:

[0020]

[0021] Wherein, a2 is the length of the line segment between the third marking point C (x3, y3) and the fourth marking point D (x4, y4), a1 is the length of the line segment between the first marking point A (x1, y1) and the second marking point B (x2, y2), and l is the precision error ratio allowed by scaling.

[0022] Preferably, the calculation formula of the rotation parameter is:

[0023] θ=θ0+Ω

[0024] Wherein θ is the rotation parameter, θ0 is the original rotation parameter in the original correction parameter, and Ω is the angle between the line segment AB connecting the first marking point A (x1, y1) and the second marking point B (x2, y2) and the line segment CD connecting the third marking point C (x3, y3) and the fourth marking point D (x4, y4).

[0025] Preferably, the angle between the line segment AB and the line segment CD is calculated as follows:

[0026]

[0027] Preferably, the translation parameters include a translation parameter c1 along the x-axis and a translation parameter c2 along the y-axis. The calculation formula of the translation parameter c1 along the x-axis is:

[0028] c1=c x +Δd i

[0029] where c x is the horizontal coordinate of the anchor point coordinate in the affine transformation matrix, Δd i is the distance increment at the i-th iteration. When translating along the x-axis, the third marker point C (x3, y3) and the fourth marker point D (x4, y4) are iteratively made to approach half the distance between the first marker point A (x1, y1) and the second marker point B (x2, y2) in the x-axis direction each time. After iteration, the third marker point C (x3, y3) and the fourth marker point D (x4, y4) are passed through the affine transformation matrix to adjust the position of the anchor point coordinates in the affine transformation matrix. If the coordinates of the third marker point C (x3, y3) and the fourth marker point D (x4, y4) after affine transformation are equal to the coordinates of the first marker point A (x1, y1) and the second marker point B (x2, y2), the c1 calculated at this time is used as the translation parameter along the x-axis in the translation parameter;

[0030] Correspondingly, the translation parameters along the y-axis in the translation parameters are calculated.

[0031] Preferably, the distance increment Δd at the i-th iteration is i The expression is,

[0032]

[0033] Beneficial effects of the present invention: Due to the adoption of the above technical solution, the present invention does not need to erect a correction rod, and obtains new correction parameters through image comparison and algorithm verification, thereby meeting the requirements of video rendering. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 Schematic diagram of the steps of a method for correcting spatial consistency of multi-view surround shooting according to an embodiment of the present invention;

[0035] Figure 2 This is a schematic diagram of step S3 in an embodiment of the present invention;

[0036] Figure 3 Schematic diagram of scene arrangement of camera positions in an embodiment of the present invention;

[0037] Figure 4 Schematic diagram of camera jitter in an embodiment of the present invention

[0038] Figure 5 A schematic diagram of marking a background image during zooming in an embodiment of the present invention;

[0039] Figure 6 Schematic diagram of marking the foreground image during zooming in an embodiment of the present invention;

[0040] Figure 7 Schematic diagram of marking of the background image during rotation according to an embodiment of the present invention;

[0041] Figure 8 Schematic diagram of marking of the foreground image during rotation according to an embodiment of the present invention;

[0042] Figure 9 Schematic diagram of marking of the background image during translation in an embodiment of the present invention;

[0043] Figure 10 Schematic diagram of marking of the foreground image during translation according to an embodiment of the present invention;

[0044] Figure 11 Schematic diagram of the contents of the correction file obtained in an embodiment of the present invention. DETAILED DESCRIPTION

[0045] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making any creative efforts shall fall within the scope of protection of the present invention.

[0046] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features in the embodiments may be combined with each other.

[0047] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but they are not intended to limit the present invention.

[0048] A correction method for spatial consistency of multi-view surround shooting, such as Figure 1 Shown, including,

[0049] Step S1: extracting frames from N camera positions to obtain a sequence frame file, and correcting the N camera positions in the sequence frame file based on a reference camera position to obtain an initial correction file and initial correction parameters for each camera position;

[0050] Step S2: re-extract frames from the N camera positions to obtain a new sequence frame file, determine the abnormal camera position where the new offset image appears based on the playback result of the new sequence frame file, and obtain initial correction parameters for the abnormal camera position from the initial correction file;

[0051] Step S3, using the image of the abnormal camera position in the sequence frame file as the background image, and using the image of the abnormal camera position in the new sequence frame file as the foreground image, performing correction processing on the foreground image and the background image until the foreground image and the background image completely overlap, and obtaining new correction parameters based on the correction process;

[0052] In step S4, the initial correction parameters of the abnormal camera position are replaced by the new correction parameters and the image of the abnormal camera position is corrected using the new correction parameters.

[0053] Specifically, the present invention relates to a method for re-calibrating and correcting a correction file for spatial consistency, which specifically includes: obtaining a photo sequence frame file that is processed for the first time with a camera spatial consistency correction file, obtaining an AE file of a correction result of the first spatial consistency processing, re-extracting frames for the camera position that has been offset, collecting new photo sequence frame files, processing the current photo sequence frame files with the original correction file, and playing them continuously to find the offset camera position, and then adjusting parameters such as the anchor point position, deflection angle and zoom ratio, comparing the corrected video effects, and determining new correct correction parameter values.

[0054] Specifically, such as Figure 3 , Figure 4 As shown, during the actual installation and deployment of the free-viewpoint camera position 1, due to problems with the on-site environment and installation technology, there are certain differences in the actual capture of the picture, such as level, scaling, and offset. If the above situation is followed, when generating the surround video, a strong sense of picture jitter will be generated, which cannot be used for actual broadcasting. Therefore, it is necessary to correct and crop the video picture in real time during the live broadcast, and intercept the common effective area of multiple cameras to ensure that the output picture is not jittery and smooth, that is, spatial consistency processing; the picture correction technology adopted by the present invention is to solve the problems of rotation, scaling, and offset of the shooting picture on the 2D plane.

[0055] In one specific embodiment, during the preparation phase, the sequence frame files undergo a first spatial consistency check and generate an initial correction file. This initial correction file contains the scale, offset, and rotation parameters for N-1 camera positions relative to the reference camera position. Specifically, this initial correction file is used to correct the images of all camera positions during the live broadcast.

[0056] Specifically, a new frame is extracted from all current camera positions to collect a new photo sequence frame file (some cameras have already shifted). According to the parameters of the initial correction file, the current N camera position photo sequence frame files are corrected and played back continuously at slow speed. By observation, if the front and rear camera positions have uneven images, the abnormal camera position with shift is determined. Assume that the abnormal camera position is m.

[0057] Select the frame file of camera position m of the sequence frame file as the background picture, and the frame file of camera position m of the new sequence frame file as the foreground picture. The new sequence frame file adds a virtual machine position, and the frame file of the virtual machine position is the frame file of camera position m of the sequence frame file as the background; the frame file of camera position m of the new sequence frame file is the foreground.

[0058] On the background picture, mark the first marking point A (x1, y1) and the second marking point B (x2, y2); on the foreground picture, mark the third marking point C (x3, y3) and the fourth marking point D (x4, y4); it should be noted that the positions of the two marking points on the foreground picture must correspond to the positions of the two marking points on the background picture.

[0059] The correction parameters of the background image are the original correction parameters of the camera position in the initial correction file, and the correction parameters of the foreground image are the original correction parameters of the camera position in the initial correction file. The background image and the foreground image are superimposed and displayed with transparency as the parameter.

[0060] By comparing the corrected position changes of the two marked points on the background and foreground images, the estimated parameters are automatically adjusted, and iterative calculations are performed to make the two points on the background and the two points on the foreground coincide with each other in the corrected positions. After the iterative calculation is completed, the new sequence frame file is processed according to the new correction parameters and played continuously. A smooth and jittery video screen indicates that the correction is successful.

[0061] The new correction parameter value of the virtual machine position replaces the corresponding correction parameter value of the abnormal machine position in the first N machine positions, and the replaced correction parameter values of the first N machine positions are saved as the correction file of the final result.

[0062] In a preferred embodiment, Figure 2 As shown, step S3 includes,

[0063] Step S31, creating a virtual machine position, and importing the background image and the foreground image into the virtual machine position;

[0064] Step S32 , marking a first marking point A ( x1 , y1 ) and a second marking point B ( x2 , y2 ) on the background image, and marking a third marking point C ( x3 , y3 ) and a fourth marking point D ( x4 , y4 ) on the foreground image;

[0065] In step S33, the foreground image is corrected by scaling, rotating, and translating the foreground image using an affine transformation matrix so that the first marker point A (x1, y1) and the third marker point C (x3, y3) coincide with each other, and the second marker point B (x2, y2) and the fourth marker point D (x4, y4) coincide with each other.

[0066] In a preferred embodiment, the first marking point A (x1, y1) and the third marking point C (x3, y3) are located at the same position of the background image and the foreground image; the second marking point B (x2, y2) and the fourth marking point D (x4, y4) are located at the same position of the background image and the foreground image.

[0067] It should be noted that the same position of the background picture and the foreground picture mentioned in the present invention refers to the same position in the picture content. For example, a marking point is set at the upper left corner of the door frame in the picture content of the background picture. Accordingly, the marking point in the foreground picture should also be set at the upper left corner of the door frame in the picture content of the foreground picture, rather than referring to the same pixel position. Because of the camera offset, although both are the upper left corner of the door frame, the actual pixel positions on the two photos of the foreground picture and the background picture are different. The purpose of the present invention is to overlap the picture content.

[0068] In a preferred embodiment, the scaling parameters, rotation parameters, and translation parameters calculated during the correction operation in step S33 are used as new correction parameters.

[0069] Specifically, according to the spatial consistency correction method, in order to make the first marking point A (x1, y1) and the third marking point C (x3, y3) coincide with each other, and to make the second marking point B (x2, y2) and the fourth marking point D (x4, y4) coincide with each other, it is necessary to be affected by the three parameters of scaling, offset, and rotation, as well as the common effective frame of N camera positions after correction.

[0070] In the process of 4-point coincidence, the two parameters of offset and rotation are first the original correction parameters of the abnormal camera position in the AE correction file, and the scaling ratio of the line segment of the two points AB and CD is calculated first.

[0071] After obtaining the scaling factor, the rotation angle between the two segments AB and CD is calculated. Finally, CD is translated (horizontally and vertically) so that AB and CD finally coincide.

[0072] After the four points coincide, the three parameters of scale, offset, and rotation are obtained. After correction by the affine transformation matrix, the common effective frame of N camera positions is intercepted, and finally a smooth picture output after correction of N camera positions is obtained.

[0073] Affine transformation matrix:

[0074]

[0075] The known parameters in the affine transformation matrix include:

[0076] 1. Offset, that is, the anchor point coordinates (c x , c y );

[0077] 2. Rotation parameter θ

[0078] 3. Scaling parameter s

[0079] 4. Center point of the frame (hw, hh)

[0080] In a preferred embodiment, Figure 5 , Figure 6 As shown, the calculation formula of the scaling parameter is,

[0081]

[0082] Where s is the scaling parameter, s0 is the original scaling parameter in the original correction parameter, Δs is the scaling parameter increment, a3 is the length of the line segment between the third marking point C (x3, y3) and the fourth marking point D (x4, y4) after correction by the scaling parameter, and a1 is the length of the line segment between the first marking point A (x1, y1) and the second marking point B (x2, y2).

[0083] In a preferred embodiment, the scaling parameter increment is increased multiple times when scaling the foreground image. After each increase in the scaling parameter increment, it is determined whether a3 is equal to a1. If they are equal, the current scaling ratio is used as the scaling parameter. Otherwise, the scaling parameter increment is continued until a3 and a1 are equal. The scaling parameter increment is calculated as follows:

[0084]

[0085] Where a2 is the length of the line segment between the third marking point C (x3, y3) and the fourth marking point D (x4, y4), a1 is the length of the line segment between the first marking point A (x1, y1) and the second marking point B (x2, y2), and l is the accuracy error ratio allowed by scaling.

[0086] Specifically, the calculation principle of the scaling parameter is:

[0087] Calculating the foreground and background scaling ratios is to calculate the length ratio of line segment CD to line segment AB.

[0088] formula:

[0089] t1=x2-x1;

[0090] t2=y2-y1

[0091] a1=sqrt(t1*t1+t2*t2);

[0092] t3 = x4 - x4;

[0093] t4=y4-y3;

[0094] a2=sqrt(t3*t3+t4*t4);

[0095] a1 is the length of line segment AB, and a2 is the length of line segment CD.

[0096] Due to the influence of the common effective frame of N camera positions, the scaling algorithm cannot simply set the scaling parameters s of the foreground and background to be directly equal to a2 / a1.

[0097] In this embodiment, l is specifically taken as 0.1, so,

[0098]

[0099] According to the affine transformation formula, the rotation parameters and anchor point coordinates are the parameters of the camera position in the initial correction file, the initial value of the scaling parameter is the initial scaling parameter of the abnormal camera position in the initial correction file, and each increase of Δs is s=s0+Δs. The coordinates of points C and D after the foreground transformation C'(x5, y5) and D'(x6, y6) are calculated, and a3 is calculated according to the length formula of the line segment C'D'.

[0100] If a3 is equal to a1, the calculation is complete, and s is the final corrected scaling parameter. If a3 is not equal to a1, continue to increase Δs and calculate the coordinates C' and D' of the corrected foreground image according to the affine transformation matrix until a3 and a1 are equal.

[0101] In a preferred embodiment, Figure 7 , Figure 8 As shown, the calculation formula of the rotation parameter is,

[0102] θ=θ0+Ω

[0103] Where θ is the rotation parameter, θ0 is the original rotation parameter in the original correction parameter, and Ω is the angle between the line segment AB connecting the first marking point A (x1, y1) and the second marking point B (x2, y2) and the line segment CD connecting the third marking point C (x3, y3) and the fourth marking point D (x4, y4).

[0104] In a preferred embodiment, the angle between line segment AB and line segment CD is calculated as follows:

[0105]

[0106] Figure 5 middle,

[0107]

[0108] Figure 6 middle,

[0109]

[0110] Specifically, the principle of rotation calculation is to calculate the rotation angle of the foreground and background, that is, to calculate the angle between the CD segment and the AB segment. Ω is the angle between the foreground and the background. Since the angle conversion is not affected by the common effective frame of N camera positions, the initial correction file is the initial rotation parameter of the camera position plus the angle between the AB segment and the CD segment.

[0111] θ=θ0+Ω

[0112] θ is the final corrected rotation angle parameter.

[0113] In a preferred embodiment, Figure 9 , Figure 10 As shown, the translation parameters include the translation parameter c1 along the x-axis and the translation parameter c2 along the y-axis. The calculation formula of the translation parameter c1 along the x-axis is:

[0114] c1=c x +Δd i

[0115] where c x is the horizontal coordinate of the anchor point coordinate in the affine transformation matrix, Δd i is the distance increment at the i-th iteration. When translating along the x-axis, the third marker point C (x3, y3) and the fourth marker point D (x4, y4) are iteratively made to approach half the distance of the first marker point A (x1, y1) and the second marker point B (x2, y2) in the x-axis direction each time. After iteration, the third marker point C (x3, y3) and the fourth marker point D (x4, y4) are passed through the affine transformation matrix to adjust the position of the anchor point coordinates in the affine transformation matrix. If the coordinates of the third marker point C (x3, y3) and the fourth marker point D (x4, y4) after affine transformation are equal to the coordinates of the first marker point A (x1, y1) and the second marker point B (x2, y2), the c1 calculated at this time is used as the translation parameter along the x-axis in the translation parameter;

[0116] The translation parameters along the y-axis are calculated accordingly.

[0117] In a preferred embodiment, the distance increment Δd at the i-th iteration is i The expression is,

[0118]

[0119] Specifically, after scaling and rotation, the third marker point C (x3, y3) and the fourth marker point D (x4, y4) are translated along the x-axis and y-axis respectively so that the first marker point A (x1, y1) and the third marker point C (x3, y3) coincide with each other, and the second marker point B (x2, y2) and the fourth marker point D (x4, y4) coincide with each other.

[0120] Specifically, after calculating the scaling and rotation angle parameters between the foreground and background, the line segments AB and CD in the foreground and background are parallel and of the same length. Finally, the third marker point C (x3, y3) and the fourth marker point D (x4, y4) need to be translated along the x-axis and y-axis so that the coordinates of the third marker point C (x3, y3) and the fourth marker point D (x4, y4) are equal to the coordinates of the first marker point A (x1, y1) and the second marker point B (x2, y2).

[0121] At this point, all calculations of the corrected scaling, rotation, and translation are completed, and finally the foreground image and the background image completely overlap after being corrected using the scaling parameters, rotation parameters, and translation parameters as new correction parameters.

[0122] In a specific embodiment, after the on-site construction is completed, the first frame is collected, and each camera position draws one frame to form a sequence frame. The three data of each camera position are obtained through the correction program: offset, rotation and scaling parameters. The content of the obtained correction file is as follows: Figure 11 As shown:

[0123] Since the video images provided by each camera position must undergo correction and other processing, in order to ensure that the rendering time can meet the efficiency requirements of live broadcast, the present invention uses the CUDA engine to realize the rotation, offset and scaling of each frame of each camera position according to the correction parameters.

[0124] The above description is only a preferred embodiment of the present invention and does not limit the implementation mode and protection scope of the present invention. For those skilled in the art, it should be aware that all solutions obtained by equivalent substitutions and obvious changes made using the description and illustrations of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for correcting spatial consistency of multi-view surround shooting, characterized in that: include, Step S1: In the shooting preparation stage, frames of images from N camera positions are extracted to obtain a sequence frame file, and the images of the N camera positions in the sequence frame file are corrected based on an image of a reference camera position to obtain an initial correction file and initial correction parameters for each camera position; Step S2: During the actual shooting phase, re-extract frames from the N camera positions to obtain a new sequence frame file, determine the abnormal camera position where the new offset image appears based on the playback results of the new sequence frame file, and obtain initial correction parameters for the abnormal camera position from the initial correction file; Step S3, using the image of the camera position corresponding to the abnormal camera position in the sequence frame file as the background image, and using the image of the abnormal camera position in the new sequence frame file as the foreground image, performing correction processing on the foreground image and the background image until the foreground image and the background image completely overlap, and obtaining new correction parameters based on the correction processing process; Step S4: replacing the initial correction parameters of the abnormal camera position with the new correction parameters and correcting the image of the abnormal camera position using the new correction parameters.

2. The method for correcting spatial consistency of multi-view surround shooting according to claim 1, characterized in that: The step S3 includes: Step S31, establishing a virtual machine position, and importing the background image and the foreground image into the virtual machine position; Step S32: Mark a first mark point on the background image and the second marker , mark a third mark point on the foreground image and the fourth marker ; Step S33, scaling, rotating, and translating the foreground image using an affine transformation matrix to correct the first marker point and the third marking point Coincidentally make the second mark point and the fourth marking point coincide.

3. The method for correcting spatial consistency of multi-view surround shooting according to claim 2, wherein: The first marking point and the third marking point Located at the same position of the background image and the foreground image; the second marking point and the fourth marking point Located at the same position of the background image and the foreground image.

4. The method for correcting spatial consistency of multi-view surround shooting according to claim 2, wherein: The scaling parameters, rotation parameters, and translation parameters calculated during the correction operation in step S33 are used as the new correction parameters.

5. The method for correcting spatial consistency of multi-view surround shooting according to claim 4, characterized in that: The calculation formula of the scaling parameter is: ； in is the scaling parameter, is the original scaling parameter in the initial correction parameter, is the scaling parameter increment, is the third marker point corrected by the scaling parameter and the fourth marking point The length of the line segment between them, The first marking point and the second marking point The length of the line segment connecting them.

6. The method for correcting spatial consistency of multi-view surround shooting according to claim 5, characterized in that: When scaling the foreground image, the scaling parameter increment is increased multiple times, and after each increase of the scaling parameter increment, a determination is made. Is it equal to If they are equal, the current scaling ratio is used as the scaling parameter, otherwise the scaling parameter increment is increased until and Until they are equal, the calculation formula for the scaling parameter increment is: ； in The third marking point and the fourth marking point The length of the line segment between them, The first marking point and the second marking point The length of the line segment between them, The precision error ratio allowed for scaling.

7. The method for correcting spatial consistency of multi-view surround shooting according to claim 4, wherein: The calculation formula of the rotation parameter is: ； in is the rotation parameter, is the original rotation parameter in the initial correction parameter, The first marking point and the second marking point The line segment AB between the two points and the fourth marking point The angle between the line segment CD connecting them.

8. The method for correcting spatial consistency of multi-view surround shooting according to claim 7, wherein: The angle between the line segment AB and the line segment CD is calculated as follows: 。 9. The method for correcting spatial consistency of multi-view surround shooting according to claim 4, wherein: The translation parameters include the translation parameters along the x-axis and the translation parameter along the y-axis , the translation parameter along the x-axis The calculation formula is, ； in is the horizontal coordinate of the anchor point coordinate in the affine transformation matrix, is the distance increment at the i-th iteration. When translating along the x-axis, the third marker point is iteratively moved and the fourth marking point Each time approaching the first marking point in the x-axis direction and the second marking point Half the distance, after iteration the third marker point and the fourth marking point By using the affine transformation matrix, the position of the anchor point coordinates in the affine transformation matrix is adjusted. If the third marker point after affine transformation is and the fourth marking point The coordinates of the first marked point and the second marking point The coordinates of are equal, and the calculated As the translation parameter along the x-axis in the translation parameter; Correspondingly, the translation parameters along the y-axis in the translation parameters are calculated.

10. The method for correcting spatial consistency of multi-view surround shooting according to claim 9, wherein: The distance increment at the i-th iteration The expression is, 。

Images