Video processing method and device and storage medium

Abstract

The invention discloses a video processing method and device and a storage medium, and belongs to the field of image processing. The method comprises the steps that a first video is acquired based on an original video, the original video comprises multiple frames of pictures obtained by shooting a target object from multiple visual angles, the first video comprises multiple frames of first object images, and the multiple frames of first object images are images of the target object in the multiple frames of pictures; the appearance styles of multiple frames of first object images included in the first video are adjusted, at least one second video corresponding to at least one appearance style is obtained, each second video comprises multiple frames of second object images, and the appearance styles of the multiple frames of second object images included in each second video are the appearance styles corresponding to the second videos; and obtaining at least one first three-dimensional Gaussian model based on the original video and the at least one second video, the appearance style of the at least one first three-dimensional Gaussian model being the same as the at least one appearance style. According to the invention, the appearance of the three-dimensional Gaussian model can be enriched.

Description

Technical Field

[0001] The present application relates to the field of image processing, and in particular to a method, device and storage medium for processing a video. Background Art

[0002] 3D Gaussian technology is a technology used for 3D scene representation and rendering. Through 3D Gaussian technology, real-world objects can be reconstructed 1:1 to obtain a 3D Gaussian model with the same shape as the real-world object.

[0003] In the related art, a three-dimensional Gaussian model reconstructed using three-dimensional Gaussian technology is identical to an object in the real world, resulting in the appearance of the reconstructed three-dimensional Gaussian model being too monotonous. Summary of the invention

[0004] The present application provides a method, device and storage medium for processing a video to enrich the appearance of a three-dimensional Gaussian model. The technical solution is as follows: In a first aspect, the present application provides a method for processing a video, in which a first video is obtained based on an original video, the original video includes multiple frames of pictures obtained by shooting a target object from multiple perspectives, and the first video includes multiple frames of first object images, which are images of the target object in the multiple frames of pictures. The appearance style of the multiple frames of first object images included in the first video is adjusted to obtain at least one second video corresponding to at least one appearance style, each second video includes multiple frames of second object images, and the appearance styles of the multiple frames of second object images included in each second video are the appearance styles corresponding to each second video. Based on the original video and at least one second video, at least one first three-dimensional Gaussian model is obtained, and the appearance style of at least one first three-dimensional Gaussian model is the same as at least one appearance style.

[0005] By adjusting the appearance styles of the multiple frames of the first object images included in the first video, at least one second video corresponding to at least one appearance style is obtained, so that at least one first three-dimensional Gaussian model with different appearance styles is obtained based on the original video and at least one second video. In other words, one or more first three-dimensional Gaussian models with different appearance styles can be constructed, thereby enriching the appearance and style of the three-dimensional Gaussian model.

[0006] In one possible implementation, at least one appearance style includes a target appearance style. Based on a target filter template, the parameter value of at least one illumination parameter of each frame of the first object image included in the first video is adjusted to obtain a second video corresponding to the target appearance style, and the target filter template is used to save the standard parameter value of at least one illumination parameter, and the standard parameter value is used to indicate the target appearance style. Alternatively, based on an image style modification model and the first video, the appearance styles of multiple frames of the first object image included in the first video are modified to a target appearance style to obtain a second video corresponding to the target appearance style. In this way, second videos with different appearance styles are obtained by using filter templates or image style modification models with different appearance styles, and three-dimensional Gaussian models with different appearance styles can be constructed based on second videos with different appearance styles.

[0007] In another possible implementation, at least one appearance style includes a target appearance style, and the second video corresponding to the target appearance style includes multiple frames of second object images corresponding to the multiple frames of pictures. The camera pose and camera parameters corresponding to each frame of the original video are obtained. Based on the camera pose, camera parameters and second object image corresponding to each frame of the picture, a first three-dimensional Gaussian model of the target appearance style is constructed. Since the second object image corresponding to each frame of the picture is introduced when constructing the three-dimensional Gaussian model, and the appearance style of the second object image corresponding to each frame of the picture is the target appearance style, a three-dimensional Gaussian model of the target appearance style can be constructed.

[0008] In another possible implementation, at least one appearance style includes a target appearance style; a second three-dimensional Gaussian model of the target object is obtained based on the original video. A parameter value of at least one illumination parameter of the second three-dimensional Gaussian model is adjusted to obtain a third three-dimensional Gaussian model. When the difference between the third three-dimensional Gaussian model and the second video corresponding to the target appearance style satisfies a difference condition, the second three-dimensional Gaussian model and the third three-dimensional Gaussian model are superimposed to obtain a first three-dimensional Gaussian model of the target appearance style. This enriches the way of obtaining a three-dimensional Gaussian model of the target appearance style.

[0009] In another possible implementation, when the difference between the third three-dimensional Gaussian model and the second video corresponding to the target appearance style does not meet the difference condition, the parameter value of at least one illumination parameter of the third three-dimensional Gaussian model is adjusted to obtain a fourth three-dimensional Gaussian model. When the difference between the fourth three-dimensional Gaussian model and the second video corresponding to the target appearance style meets the difference condition, the second three-dimensional Gaussian model and the fourth three-dimensional Gaussian model are superimposed to obtain a first three-dimensional Gaussian model of the target appearance style. In this way, a three-dimensional Gaussian model of the target appearance style can be accurately obtained.

[0010] In another possible implementation manner, the at least one illumination parameter of the second three-dimensional Gaussian model includes one or more of the following: color or transparency.

[0011] In another possible implementation, the correspondence between at least one third 3D Gaussian model and the second 3D Gaussian model is saved, at least one third 3D Gaussian model corresponds to at least one second video, and the difference between each third 3D Gaussian model and the second video corresponding to each third 3D Gaussian model satisfies the difference condition. In this way, the storage resources required for saving can be reduced.

[0012] In a second aspect, the present application provides a device for processing a video, the device comprising: an acquisition unit and an adjustment unit.

[0013] an acquisition unit, configured to acquire a first video based on an original video, wherein the original video includes a plurality of frames of pictures obtained by photographing a target object from a plurality of perspectives, and the first video includes a plurality of frames of first object images, wherein the plurality of frames of first object images are images of the target object in the plurality of frames of pictures; An adjustment unit, configured to adjust the appearance styles of the multiple frames of first object images included in the first video to obtain at least one second video corresponding to at least one appearance style, each second video including multiple frames of second object images, and the appearance styles of the multiple frames of second object images included in each second video are respectively the appearance styles corresponding to each second video; The acquisition unit is further used to acquire at least one first three-dimensional Gaussian model based on the original video and at least one second video, and the appearance style of the at least one first three-dimensional Gaussian model is the same as the at least one appearance style.

[0014] Since the adjustment unit adjusts the appearance styles of the multiple frames of first object images included in the first video, at least one second video corresponding to at least one appearance style is obtained, so that the acquisition unit acquires at least one first three-dimensional Gaussian model with different appearance styles based on the original video and at least one second video. In other words, one or more first three-dimensional Gaussian models with different appearance styles can be constructed, thereby enriching the appearance and style of the three-dimensional Gaussian model.

[0015] In a possible implementation, at least one appearance style includes a target appearance style. The adjustment unit is configured to: Based on the target filter template, adjust the parameter value of at least one illumination parameter of each frame of the first object image included in the first video to obtain a second video corresponding to the target appearance style, and the target filter template is used to store the standard parameter value of at least one illumination parameter, and the standard parameter value is used to indicate the target appearance style. Or, Based on the image style modification model and the first video, the appearance styles of multiple frames of first object images included in the first video are modified to a target appearance style, so as to obtain a second video corresponding to the target appearance style.

[0016] In this way, second videos with different appearance styles can be obtained through filter templates or image style modification models with different appearance styles, and three-dimensional Gaussian models with different appearance styles can be constructed based on the second videos with different appearance styles.

[0017] In another possible implementation, at least one appearance style includes a target appearance style, and the second video corresponding to the target appearance style includes multiple frames of second object images corresponding to the multiple frames of pictures. The acquisition unit is used to: Get the camera pose and camera parameters corresponding to each frame of the original video; Based on the camera pose, camera parameters and the second object image corresponding to each frame of the image, a first three-dimensional Gaussian model of the target appearance style is constructed.

[0018] Since the second object image corresponding to each frame is introduced when constructing the three-dimensional Gaussian model, and the appearance style of the second object image corresponding to each frame is the target appearance style, a three-dimensional Gaussian model of the target appearance style can be constructed.

[0019] In another possible implementation, the at least one appearance style includes a target appearance style; The acquisition unit is further used to acquire a second three-dimensional Gaussian model of the target object based on the original video; The adjustment unit is further used to adjust a parameter value of at least one illumination parameter of the second three-dimensional Gaussian model to obtain a third three-dimensional Gaussian model; The acquisition unit is used to superimpose the second 3D Gaussian model and the third 3D Gaussian model to obtain the first 3D Gaussian model of the target appearance style when the difference between the third 3D Gaussian model and the second video corresponding to the target appearance style meets the difference condition. This enriches the way of obtaining the 3D Gaussian model of the target appearance style.

[0020] In another possible implementation, the adjustment unit is further used to adjust the parameter value of at least one illumination parameter of the third three-dimensional Gaussian model to obtain a fourth three-dimensional Gaussian model when the difference between the third three-dimensional Gaussian model and the second video corresponding to the target appearance style does not meet the difference condition.

[0021] The acquisition unit is further configured to superimpose the second three-dimensional Gaussian model and the fourth three-dimensional Gaussian model to obtain a first three-dimensional Gaussian model of the target appearance style when the difference between the fourth three-dimensional Gaussian model and the second video corresponding to the target appearance style satisfies a difference condition. In this way, the three-dimensional Gaussian model of the target appearance style can be accurately obtained.

[0022] In another possible implementation manner, the at least one illumination parameter of the second three-dimensional Gaussian model includes one or more of the following: color or transparency.

[0023] In another possible implementation, the device further includes a storage unit, which is used to store the correspondence between at least one third three-dimensional Gaussian model and the second three-dimensional Gaussian model, at least one third three-dimensional Gaussian model corresponds to at least one second video, and the difference between each third three-dimensional Gaussian model and the second video corresponding to each third three-dimensional Gaussian model satisfies the difference condition. In this way, the storage resources required for storage can be reduced.

[0024] In a third aspect, the present application provides a computing device cluster, the computing device cluster comprising at least one computing device, each computing device comprising a processor and a memory; The processor of the at least one computing device is used to execute instructions stored in the memory of the at least one computing device, so that the computing device cluster executes the method in the first aspect or any possible implementation manner of the first aspect.

[0025] In a fourth aspect, the present application provides a computer program product comprising instructions, which, when executed by a computing device cluster, causes the computing device cluster to execute the method in the first aspect or any possible implementation manner of the first aspect.

[0026] In a fifth aspect, the present application provides a computer-readable storage medium, comprising computer program instructions. When the computer program instructions are executed by a computing device cluster, the computing device cluster executes the method in the first aspect or any possible implementation of the first aspect.

[0027] In a sixth aspect, the present application provides a chip comprising a memory and a processor, the memory being used to store computer instructions, and the processor being used to call and run the computer instructions from the memory to execute the method in the above-mentioned first aspect or any possible implementation of the first aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 is a schematic diagram of a target object provided in an embodiment of the present application; Figure 2 is a schematic diagram of an original video provided in an embodiment of the present application; Figure 3 is a schematic diagram of a first video provided in an embodiment of the present application; Figure 4 is a schematic diagram of a three-dimensional Gaussian model provided in an embodiment of the present application; Figure 5 is a schematic diagram of the structure of a computing device provided in an embodiment of the present application; Figure 6 is a flow chart of a method for processing a video provided by an embodiment of the present application; Figure 7is a flow chart of another method for processing a video provided in an embodiment of the present application; Figure 8 is a schematic diagram of the structure of a device for processing video provided in an embodiment of the present application; Fig. 9 is a schematic diagram of the structure of another computing device provided in an embodiment of the present application; Fig.10 is a schematic diagram of a computing device cluster structure provided in an embodiment of the present application; Fig.11 This is another schematic diagram of a computing device cluster structure provided in an embodiment of the present application. DETAILED DESCRIPTION

[0029] A camera can be used to shoot a target object from multiple different perspectives to obtain an original video, which includes multiple frames of images, each of which has a different perspective. Each frame of the original image includes an image of the target object and a background image of the environment in which the target object is located. Since each frame of the image has a different perspective, each frame of the image includes images of different sides of the target object.

[0030] For example, see Figure 1 The target object shown is shot from five different perspectives to obtain the original video, where the five perspectives are the front perspective, the right side perspective, the rear perspective, the left side perspective and the top perspective, and the original video includes five frames of pictures.

[0031] See also Figure 2 , use the camera to shoot the front of the target object from the front view to obtain picture a, which includes the front image and background image of the target object, that is, Figure 2 The right side of the target object is photographed by a camera from the right side view to obtain picture b, which includes the right side image of the target object, namely Figure 2 The image b in the figure is taken from the back of the target object using a camera. The image c is obtained by shooting the back of the target object from the perspective of the back. The image c includes the back image of the target object and the background image, which is Figure 2 The left side of the target object is photographed from the left side using a camera to obtain image d, which includes the left side image of the target object, namely Figure 2 The top surface of the target object is photographed by a camera from a top-down perspective to obtain image e, which includes the top surface image of the target object and the background image, namely Figure 2 Thus, the original video includes five frames of pictures, that is, the original video includes picture a, picture b, picture c, picture d and picture e.

[0032] A first video is obtained based on an original video, the first video includes multiple frames of object images, and the multiple frames of object images are images of target objects in multiple frames of pictures. The multiple frames of object images correspond to the multiple frames of pictures one by one, and the camera pose and camera parameters corresponding to each frame of pictures are obtained. Based on the camera pose, camera parameters and target object images corresponding to each frame of pictures, a three-dimensional Gaussian model with the same shape, color and texture as the target object is constructed.

[0033] For example, based on Figure 2 The original video shown includes five frames of pictures to obtain a first video, and the first video includes five frames of object images. Figure 2 The original video shown includes picture a, from Figure 2 The front image of the target object is obtained from the image a in Figure 3 As shown in Figure (1) in the figure. Figure 2 The original video shown includes picture b, from Figure 2 The right side image of the target object is obtained from the picture b in Figure 3 As shown in Figure (2) in the figure. Figure 2 The original video shown includes picture c, from Figure 2 The back image of the target object is obtained from the image c in Figure 3 As shown in Figure (3) in the figure. Figure 2 The original video shown includes picture d, from Figure 2 The left side image of the target object is obtained from the image d in Figure 3 As shown in Figure (4) in the figure. Figure 2 The original video shown includes picture e, from Figure 2 The top surface image of the target object is obtained from the image e in Figure 3 As shown in Figure (5) in . Therefore, the first video obtained includes five frames of object images, which are the front image, right side image, back image, left side image and top image of the target object. The original pictures include picture a, picture b, picture c, picture d and picture e, which correspond to the front image, right side image, back image, left side image and top image of the target object included in the first video.

[0034] The original video includes the camera pose and camera parameters corresponding to picture a, the camera pose and camera parameters corresponding to picture b, the camera pose and camera parameters corresponding to picture c, the camera pose and camera parameters corresponding to picture d, and the camera pose and camera parameters corresponding to picture e. Based on the camera pose, camera parameters, and front image corresponding to picture a, the camera pose, camera parameters, and right side image corresponding to picture b, the camera pose, camera parameters, and back image corresponding to picture c, the camera pose, camera parameters, and left side image corresponding to picture d, and the camera pose, camera parameters, and top image corresponding to picture e, a three-dimensional Gaussian model with the same shape, color, and texture as the target object is constructed. The three-dimensional Gaussian model is as follows: Figure 4 shown.

[0035] Since the constructed 3D Gaussian model is the same as the target object in the real world, the appearance of the 3D Gaussian model is too simple and not beautiful enough. In order to enrich the appearance of the 3D Gaussian model and improve the aesthetics of the 3D Gaussian model, any of the following embodiments can be used to process the video to obtain at least one 3D Gaussian model with different appearance styles.

[0036] See also Figure 5 An embodiment of the present application provides a computing device 500 , which includes a material processor 501 , a material equalizer 502 , and a model building module 503 .

[0037] The material processor 501 is used to obtain an original video, which includes multiple frames of pictures obtained by shooting a target object from multiple perspectives, and obtain a first video based on the original video. The first video includes multiple frames of first object images, which are images of the target object in the multiple frames of pictures.

[0038] The material equalizer 502 is used to adjust the appearance style of multiple frames of first object images included in the first video to obtain at least one second video corresponding to at least one appearance style, each second video includes multiple frames of second object images, and the appearance styles of the multiple frames of second object images included in each second video are the appearance styles corresponding to each second video, and the appearance styles of the multiple frames of second object images included in each second video are different.

[0039] The model building module 503 is used to obtain at least one first three-dimensional Gaussian model of the target object based on the original video and at least one second video, the appearance style of at least one first three-dimensional Gaussian model is the same as at least one appearance style, and the appearance style of each first three-dimensional Gaussian model is different.

[0040] In some embodiments, see Figure 5 , the computing device 500 also includes a lighting parser 504; The illumination analyzer 504 is used to analyze the illumination condition of each frame of the original video to obtain a parameter value of at least one illumination parameter of each frame of the original video; The material equalizer 502 is used to adjust the appearance style of multiple frames of first object images included in the first video based on the parameter value of at least one illumination parameter of each frame of the original video to obtain at least one second video corresponding to at least one appearance style.

[0041] In some embodiments, see Figure 5 The computing device 500 also includes a camera estimation module 505; at least one appearance style includes a target appearance style, and the multiple frames of second object images included in the second video corresponding to the target appearance style correspond one to one with the multiple frames of pictures included in the original video.

[0042] The camera estimation module 505 is used to obtain the camera pose and camera parameters corresponding to each frame of the original video.

[0043] The model building module 503 is used to build a first three-dimensional Gaussian model of the target appearance style based on the camera pose, camera parameters and the second object image corresponding to each frame of the image.

[0044] In some embodiments, the method in which the computing device 500 processes the original video to obtain at least one first three-dimensional Gaussian model of the target object can be applied to one or more of the following scenarios: 1. Any media application scenario; 2. Game scene; 3. Advertising, publicity, film and television video production scenes; 4. Any display scenarios such as cultural tourism, e-commerce, and corporate exhibition halls; 5. Extended Reality (XR) glasses.

[0045] In the embodiment of the present application, since the material equalizer 502 adjusts the appearance styles of the multiple frames of the first object images included in the first video, at least one second video corresponding to at least one appearance style is obtained, and the appearance styles of each second video are different, the model construction module constructs at least one first three-dimensional Gaussian model of at least one appearance style based on the original video and at least one second video. In other words, one or more first three-dimensional Gaussian models with different appearance styles can be constructed, thereby enriching the style of the three-dimensional Gaussian model and improving the aesthetics of the three-dimensional Gaussian model.

[0046] See also Figure 6 The present application embodiment provides a method 600 for processing a video, which can be applied to Figure 5The computing device shown, the method 600 obtains a first video based on an original video, adjusts the appearance style of multiple frames of first object images included in the first video, and obtains at least one second video corresponding to at least one appearance style, and the multiple frames of second object images included in each second video correspond to the multiple frames of pictures included in the original video. At least one first three-dimensional object model of at least one appearance style is constructed based on the camera pose, parameters and second object image corresponding to each frame of the picture. The method 600 includes the following process.

[0047] Step 601: Acquire a first video based on an original video, wherein the original video includes multiple frames of pictures obtained by photographing a target object from multiple perspectives, and the first video includes multiple frames of first object images, which are images of the target object in the multiple frames of pictures.

[0048] A camera may be used to photograph different surfaces of a target object to obtain multiple frames of images obtained by photographing the target object from multiple perspectives, that is, to obtain an original video including the multiple frames of images.

[0049] For each frame of the original video, a first object image of the target object is extracted from the image. The first object image of the target object is extracted from each frame of the image in the same manner as above, and multiple frames of first object images corresponding to the multiple frames of the image are obtained, so as to obtain a first video including the multiple frames of first object images.

[0050] In some embodiments, the computing device includes a material processor, and the material processor is used to extract the first object image in each frame of the original video to obtain the second video.

[0051] Step 602: Adjust the appearance styles of the multiple frames of first object images included in the first video to obtain at least one second video corresponding to at least one appearance style, each second video includes multiple frames of second object images, and the appearance styles of the multiple frames of second object images included in each second video are the appearance styles corresponding to each second video.

[0052] At least one appearance style includes a target appearance style. For each frame of the first object image in the first video, the appearance style of the frame of the first object image is adjusted to the target appearance style to obtain a frame of the second object image of the target appearance style. In the same manner as above, the appearance style of each frame of the first object image is adjusted to the target appearance style to obtain multiple frames of the second object image of the target appearance style, so as to obtain a second video including the multiple frames of the second object image, that is, to obtain a second video corresponding to the target appearance style. At least one appearance style also includes other appearance styles. The same operation as above can be used to obtain second videos corresponding to other appearance styles.

[0053] In some embodiments, the computing device includes at least one filter template, the at least one filter template corresponds to at least one appearance style, and each filter template is used to indicate the appearance style corresponding to each filter template.

[0054] For each filter template included in the at least one filter template, the filter template includes a standard parameter value of at least one illumination parameter, and the standard parameter value of the at least one illumination parameter is used to indicate an appearance style corresponding to the filter template.

[0055] The standard parameter values ​​of the illumination parameters included in the filter template corresponding to each appearance style may be different, so that each filter template is used to indicate the appearance style corresponding to each filter template.

[0056] For example, at least one appearance style includes one or more of the following: Ghibli style, sparkling water style, or long summer style, etc. The standard parameter value of at least one illumination parameter included in the filter template corresponding to the Ghibli style, the standard parameter value of at least one illumination parameter included in the filter template corresponding to the sparkling water style, and the standard parameter value of at least one illumination parameter included in the filter template corresponding to the long summer style are different.

[0057] The at least one lighting parameter includes one or more of the following: lighting matrix, color temperature, brightness, sharpness, saturation or white balance, etc.

[0058] For each filter template included in at least one filter template, for the sake of convenience of explanation, the filter template is referred to as a target filter template, and the appearance style corresponding to the target filter template is referred to as a target appearance style. In step 602, based on the target filter template, the parameter value of at least one illumination parameter of each frame of the first object image included in the first video is adjusted to obtain a second video corresponding to the target appearance style.

[0059] Optionally, during implementation, the second video corresponding to the target appearance style may be obtained according to the following process 6021-6023.

[0060] 6021: Obtain a parameter value of at least one illumination parameter of each frame of the original video, and obtain an average parameter value of the at least one illumination parameter based on the parameter value of the at least one illumination parameter of each frame of the original video.

[0061] The computing device includes an illumination parser and a material equalizer, and the original video can be input into the illumination parser, and the illumination condition of each frame of the original video is parsed by the illumination parser to obtain a parameter value of at least one illumination parameter of each frame of the original video. Based on the parameter value of at least one illumination parameter of each frame of the original video, an average parameter value of the at least one illumination parameter is obtained. Next, the material equalizer is used to adjust the appearance style of the first video according to the following process to obtain a second video corresponding to the target appearance style.

[0062] 6022: Obtain an offset value of at least one illumination parameter based on a standard parameter value of at least one illumination parameter included in the target filter template and an average parameter value of the at least one illumination parameter.

[0063] For each illumination parameter of the at least one illumination parameter, an offset value of the illumination parameter is obtained based on a standard parameter value of the illumination parameter included in the target filter template and an average parameter value of the illumination parameter.

[0064] 6023: Based on the offset value of the at least one illumination parameter, adjust the first parameter value of the at least one illumination parameter of each frame of the first object image included in the first video to obtain the second parameter value of the at least one illumination parameter of each frame of the first object image. For two adjacent frames of the first object image, adjust the second parameter value of the at least one illumination parameter of the two adjacent frames of the first object image to reduce the difference between the second parameter value of the at least one illumination parameter of the two adjacent frames of the first object image to obtain two adjacent frames of the second object image.

[0065] For example, for a first frame first object image and a second frame first object image, adjust the second parameter value of at least one illumination parameter of the first frame first object image and the second parameter value of at least one illumination parameter of the second frame first object image to reduce the difference between the second parameter value of at least one illumination parameter of the first frame first object image and the second parameter value of at least one illumination parameter of the second frame first object image, and obtain a first frame second object image and a second frame second object image.

[0066] In some embodiments, for two adjacent frames of first object images, the first parameter value of at least one illumination parameter of the two adjacent frames of first object images may also be adjusted to reduce the difference between the first parameter value of at least one illumination parameter of the two adjacent frames of first object images to obtain two adjacent frames of third object images. Based on the offset value of the at least one illumination parameter, the first parameter value of at least one illumination parameter of each frame of the third object image is adjusted to obtain two adjacent frames of second object images.

[0067] In some embodiments, based on the image style modification model and the first video, the appearance styles of multiple frames of first object images included in the first video are modified to a target appearance style to obtain a second video corresponding to the target appearance style.

[0068] Optionally, the first video and the target appearance style can be input into an image style modification model, and the image style modification model modifies the multiple frames of first object images included in the first video based on the target appearance style to obtain multiple frames of second object images of the target appearance style, and outputs multiple frames of second object images of the target appearance style. In this way, a second video corresponding to the target appearance style is obtained, and the second video corresponding to the target appearance style includes multiple frames of second object images of the target appearance style.

[0069] Optionally, the image style modification model may be any model used to perform stylization processing on a video.

[0070] Step 603: Based on the original video and at least one second video, at least one first three-dimensional Gaussian model is obtained, and the appearance style of the at least one first three-dimensional Gaussian model is the same as the at least one appearance style.

[0071] The at least one first three-dimensional Gaussian model is a three-dimensional Gaussian model of the target object.

[0072] At least one second video includes a second video of a target appearance style, and based on the original video and the second video corresponding to the target appearance style, a first three-dimensional Gaussian model of the target appearance style is obtained. When implemented, it can be implemented according to the following operations 6031-6032.

[0073] 6031: Obtain the camera pose and camera parameters corresponding to each frame of the original video.

[0074] Optionally, the camera parameters include intrinsic parameters and / or extrinsic parameters of the camera.

[0075] The computing device includes a camera estimation module. For each frame of the original video, the camera estimation module is used to process the frame of the picture to obtain the camera pose and camera parameters when the camera took the frame of the picture, that is, to obtain the camera pose and camera parameters corresponding to the frame of the picture.

[0076] 6032: Based on the camera pose, camera parameters and second object image corresponding to each frame of the image, construct a first three-dimensional Gaussian model of the target appearance style.

[0077] The computing device includes a model building module, which builds a first three-dimensional Gaussian model of the target appearance style based on the camera posture, camera parameters and second object image corresponding to each frame of the picture.

[0078] In the embodiment of the present application, at least one second video corresponding to at least one appearance style is obtained, and each second video has a different appearance style, so that at least one first three-dimensional Gaussian model of at least one appearance style is obtained based on the original video and at least one second video. In other words, one or more first three-dimensional Gaussian models with different appearance styles can be obtained, thereby enriching the appearance of the three-dimensional Gaussian model.

[0079] See also Figure 7 The present application embodiment provides a method 700 for processing a video, and the method 700 can be applied to Figure 5 The computing device shown, the method 700 obtains a first video based on an original video, adjusts the appearance style of multiple frames of first object images included in the first video, and obtains at least one second video corresponding to at least one appearance style, and the multiple frames of second object images included in each second video correspond one to one to the multiple frames of pictures included in the original video. A second three-dimensional Gaussian model of the target object is constructed based on the original video, and the parameter value of at least one illumination parameter of the second three-dimensional Gaussian model is adjusted to obtain a third three-dimensional Gaussian model. When the difference between the third three-dimensional Gaussian model and the second video corresponding to the target appearance style meets the difference condition, the second three-dimensional Gaussian model and the third three-dimensional Gaussian model are superimposed to obtain a first three-dimensional Gaussian model of the target appearance style. The method 700 includes the following process.

[0080] Steps 701-702: respectively Figure 6 Steps 601 - 602 of the method 600 are the same and will not be described in detail herein.

[0081] Step 703: Acquire a second three-dimensional Gaussian model of the target object based on the original video.

[0082] In step 701, a first video is obtained based on an original video, the original video includes multiple frames of pictures obtained by shooting a target object from multiple perspectives, and the first video includes multiple frames of first object images, which are images of the target object in the multiple frames of pictures, that is, the multiple frames of first object images included in the first video correspond one-to-one to the multiple frames of pictures included in the original video.

[0083] In step 703, the camera position and camera parameters corresponding to each frame of the original video are obtained, and a second three-dimensional Gaussian model is constructed based on the camera position and camera parameters corresponding to each frame of the original video and the first object image.

[0084] The appearance style of the second three-dimensional Gaussian model is the same as the appearance style of the target object, that is, the second three-dimensional Gaussian model is a Gaussian model whose texture, color and shape are respectively the same as those of the target object.

[0085] In step 702, at least one second video corresponding to at least one appearance style is obtained. For each appearance style in the at least one appearance style, for ease of explanation, the appearance style is referred to as a target appearance style, and a first three-dimensional Gaussian model of the target appearance style is obtained according to the following process.

[0086] Step 704: Adjust the parameter value of at least one illumination parameter of the second three-dimensional Gaussian model to obtain a third three-dimensional Gaussian model.

[0087] By adjusting the parameter value of at least one illumination parameter of the second three-dimensional Gaussian model, the appearance style of the second three-dimensional Gaussian model can be changed, that is, the shape, texture and / or color of the second three-dimensional Gaussian model can be changed to obtain a third three-dimensional Gaussian model whose appearance style is different from that of the second three-dimensional Gaussian model.

[0088] Step 705: Compare whether the difference between the third 3D Gaussian model and the second video corresponding to the target appearance style meets the difference condition. If the difference between the third 3D Gaussian model and the second video corresponding to the target appearance style meets the difference condition, execute step 706; if the difference between the third 3D Gaussian model and the second video corresponding to the target appearance style does not meet the difference condition, execute step 707.

[0089] The second video includes second object images of different surfaces of the target object, and the third three-dimensional Gaussian model is composed of third object images of different surfaces of the target object. The second video includes multiple frames of second object images and the third three-dimensional Gaussian model includes multiple frames of third object images.

[0090] For each frame of the second object image and the third object image corresponding to the second object image, a peak signal-to-noise ratio (PSNR) between the second object image and the third object image is obtained. According to the above process, multiple PSNRs between multiple frames of the second object image and multiple frames of the third object image can be obtained, and an average PSNR is obtained based on the multiple PSNRs.

[0091] Among them, if the average PSNR is smaller, the difference between the third three-dimensional Gaussian model and the second video corresponding to the target appearance style is smaller, and if the average PSNR is larger, the difference between the third three-dimensional Gaussian model and the second video corresponding to the target appearance style is larger.

[0092] If the average PSNR is less than or equal to the PSNR threshold, the difference between the third three-dimensional Gaussian model and the second video corresponding to the target appearance style satisfies the difference condition; if the average PSNR is greater than the PSNR threshold, the difference between the third three-dimensional Gaussian model and the second video corresponding to the target appearance style does not satisfy the difference condition.

[0093] Step 706: superimpose the second 3D Gaussian model and the third 3D Gaussian model to obtain a first 3D Gaussian model of the target appearance style.

[0094] The second three-dimensional Gaussian model includes a plurality of first Gaussian spheres, and the third three-dimensional Gaussian model includes a plurality of second Gaussian spheres, and the plurality of first Gaussian spheres correspond one to one to the plurality of second Gaussian spheres.

[0095] In step 706, the positions, shapes, color information and transparency of the plurality of first Gaussian balls included in the second three-dimensional Gaussian model are obtained, and the positions, shapes, color information and transparency of the plurality of second Gaussian balls included in the third three-dimensional Gaussian model are obtained. Based on the positions and shapes of the plurality of first Gaussian balls and the positions and shapes of the plurality of second Gaussian balls, a plurality of Gaussian ball pairs are determined, each Gaussian ball pair includes a first Gaussian ball and a second Gaussian ball, and the position and shape of the first Gaussian ball are respectively the same as the position and shape of the second Gaussian ball. The color information of the first Gaussian ball included in each Gaussian ball pair and the color information of the second Gaussian ball are superimposed, and the transparency of the first Gaussian ball included in each Gaussian ball pair and the transparency of the second Gaussian ball are superimposed, so as to obtain the color information and transparency of the plurality of Gaussian balls included in the first three-dimensional Gaussian model of the target appearance style, that is, to obtain the first three-dimensional Gaussian model of the target appearance style.

[0096] For other appearance styles except the target appearance style in the at least one appearance style, the process returns to step 704 to start execution to obtain first three-dimensional Gaussian models of other appearance styles.

[0097] Step 707: Update the second three-dimensional Gaussian model to a third-dimensional Gaussian model, and return to execute step 704.

[0098] That is to say, continue to adjust the parameter value of at least one illumination parameter of the second three-dimensional Gaussian model until the difference between the adjusted third three-dimensional Gaussian model and the second video corresponding to the target appearance is less than a threshold value, and then superimpose the second three-dimensional Gaussian model and the adjusted third three-dimensional Gaussian model to obtain a first three-dimensional Gaussian model of the target appearance style.

[0099] At least one third three-dimensional Gaussian model corresponding to at least one second video can be obtained, and the difference between each third three-dimensional Gaussian model and the second video corresponding to each third three-dimensional Gaussian model satisfies the difference condition.

[0100] The correspondence between at least one third three-dimensional Gaussian model and the second three-dimensional Gaussian model is saved, so that when at least one first three-dimensional Gaussian model of at least one appearance style is needed, each third three-dimensional Gaussian model in the at least one third three-dimensional Gaussian model is superimposed with the second three-dimensional Gaussian model to obtain at least one first three-dimensional Gaussian model of at least one appearance style.

[0101] The total data volume of the correspondence between at least one third three-dimensional Gaussian model and the second three-dimensional Gaussian model is less than the total data of at least one first three-dimensional Gaussian model of at least one appearance style, so the correspondence between at least one third three-dimensional Gaussian model and the second three-dimensional Gaussian model is saved, which can reduce the storage resources required for saving.

[0102] In an embodiment of the present application, at least one second video corresponding to at least one appearance style is obtained, and each second video has a different appearance style, and a second three-dimensional Gaussian model is obtained based on the original video. The parameter value of at least one illumination parameter of the second three-dimensional Gaussian model is adjusted to obtain a third three-dimensional Gaussian model. The difference between the third three-dimensional Gaussian model and the second video corresponding to a certain appearance style satisfies the difference condition, and the second three-dimensional Gaussian model and the third three-dimensional Gaussian model are superimposed to obtain a first three-dimensional Gaussian model of the appearance style. In this way, at least one first three-dimensional Gaussian model corresponding to at least one appearance style can be obtained, thereby enriching the appearance of the three-dimensional Gaussian model.

[0103] See also Figure 8 The embodiment of the present application provides a device 800 for processing a video, and the device 800 can be deployed in Figure 5 on the computing device shown, or Figure 6 on a computing device as shown in method 600, or Figure 7 The computing device of the method 700 is shown. The apparatus 800 includes: an acquisition unit 801 and an adjustment unit 802.

[0104] An acquisition unit 801 is configured to acquire a first video based on an original video, where the original video includes a plurality of frames of pictures obtained by photographing a target object from a plurality of perspectives, and the first video includes a plurality of frames of first object images, where the plurality of frames of first object images are images of the target object in the plurality of frames of pictures; An adjusting unit 802 is configured to adjust the appearance styles of the multiple frames of first object images included in the first video to obtain at least one second video corresponding to at least one appearance style, each second video includes multiple frames of second object images, and the appearance styles of the multiple frames of second object images included in each second video are respectively the appearance styles corresponding to each second video; The acquisition unit 801 is further configured to acquire at least one first 3D Gaussian model based on the original video and at least one second video, wherein the appearance style of the at least one first 3D Gaussian model is the same as the at least one appearance style.

[0105] Optionally, the detailed implementation process of the acquisition unit 801 acquiring the first video based on the original video can be found in Figure 6 Step 601 of method 600 or Figure 7The relevant contents of step 701 of the method 700 are not described in detail here.

[0106] Optionally, the detailed implementation process of the adjustment unit 802 adjusting the appearance styles of the multiple frames of the first object images included in the first video can be found in Figure 6 Step 602 of method 600 or Figure 7 The relevant contents of step 702 of the method 700 are not described in detail here.

[0107] Optionally, the acquisition unit 801 acquires at least one first three-dimensional Gaussian model based on the original video and at least one second video. Figure 6 Step 603 of method 600 or Figure 7 The relevant contents of steps 703 - 707 of the method 700 are not described in detail here.

[0108] Optionally, the at least one appearance style includes a target appearance style. The adjusting unit 802 is configured to: Based on the target filter template, adjust the parameter value of at least one illumination parameter of each frame of the first object image included in the first video to obtain a second video corresponding to the target appearance style, and the target filter template is used to store the standard parameter value of at least one illumination parameter, and the standard parameter value is used to indicate the target appearance style. Or, Based on the image style modification model and the first video, the appearance styles of multiple frames of first object images included in the first video are modified to a target appearance style, so as to obtain a second video corresponding to the target appearance style.

[0109] Optionally, the adjustment unit 802 adjusts the parameter value of at least one illumination parameter of each frame of the first object image included in the first video based on the target filter template. Figure 6 Step 602 of method 600 or Figure 7 The relevant contents of step 702 of the method 700 are not described in detail here.

[0110] Optionally, the adjustment unit 802 modifies the appearance styles of the multiple frames of first object images included in the first video into the target appearance style based on the image style modification model and the first video. Figure 6 Step 602 of method 600 or Figure 7 The relevant contents of step 702 of the method 700 are not described in detail here.

[0111] Optionally, at least one appearance style includes a target appearance style, and the second video corresponding to the target appearance style includes multiple frames of second object images corresponding to the multiple frames of pictures. The acquisition unit 801 is used to: Get the camera pose and camera parameters corresponding to each frame of the original video; Based on the camera pose, camera parameters and the second object image corresponding to each frame of the image, a first three-dimensional Gaussian model of the target appearance style is constructed.

[0112] Optionally, the detailed implementation process of the acquisition unit 801 acquiring the camera pose and camera parameters corresponding to each frame of the original video is shown in Figure 6 The relevant contents of step 603 of the method 600 are not described in detail here.

[0113] Optionally, the acquisition unit 801 constructs a first three-dimensional Gaussian model of the target appearance style based on the camera pose, camera parameters and second object image corresponding to each frame of the image. Figure 6 The relevant contents of step 603 of the method 600 are not described in detail here.

[0114] Optionally, the at least one appearance style comprises a target appearance style; The acquisition unit 801 is further used to acquire a second three-dimensional Gaussian model of the target object based on the original video; The adjusting unit 802 is further configured to adjust a parameter value of at least one illumination parameter of the second three-dimensional Gaussian model to obtain a third three-dimensional Gaussian model; The acquisition unit 801 is used to superimpose the second 3D Gaussian model and the third 3D Gaussian model to obtain a first 3D Gaussian model of the target appearance style when the difference between the third 3D Gaussian model and the second video corresponding to the target appearance style meets a difference condition.

[0115] Optionally, the detailed implementation process of the acquisition unit 801 acquiring the second three-dimensional Gaussian model of the target object based on the original video can be found in Figure 7 The relevant contents of step 703 of the method 700 are not described in detail here.

[0116] Optionally, the adjustment unit 802 adjusts the parameter value of at least one illumination parameter of the second three-dimensional Gaussian model to obtain a third three-dimensional Gaussian model. Figure 7 The relevant contents of step 704 of the method 700 are not described in detail here.

[0117] Optionally, the acquisition unit 801 superimposes the second 3D Gaussian model and the third 3D Gaussian model to obtain the first 3D Gaussian model of the target appearance style. Figure 7 The relevant contents of step 706 of the method 700 are not described in detail here.

[0118] Optionally, the adjusting unit 802 is further configured to adjust a parameter value of at least one illumination parameter of the third 3D Gaussian model to obtain a fourth 3D Gaussian model when the difference between the third 3D Gaussian model and the second video corresponding to the target appearance style does not satisfy a difference condition.

[0119] The acquisition unit 801 is further configured to superimpose the second 3D Gaussian model and the fourth 3D Gaussian model to obtain a first 3D Gaussian model of the target appearance style when the difference between the fourth 3D Gaussian model and the second video corresponding to the target appearance style satisfies a difference condition.

[0120] Optionally, the adjustment unit 802 adjusts the parameter value of at least one illumination parameter of the third three-dimensional Gaussian model to obtain a fourth three-dimensional Gaussian model. Figure 7 The relevant contents of step 704 of the method 700 are not described in detail here.

[0121] Optionally, the acquisition unit 801 superimposes the second 3D Gaussian model and the fourth 3D Gaussian model to obtain the first 3D Gaussian model of the target appearance style. Figure 7 The relevant contents of step 706 of the method 700 are not described in detail here.

[0122] Optionally, the at least one illumination parameter of the second three-dimensional Gaussian model includes one or more of the following: color or transparency.

[0123] Optionally, the device 800 further includes a storage unit 803; The saving unit 803 is used to save the correspondence between at least one third three-dimensional Gaussian model and the second three-dimensional Gaussian model, at least one third three-dimensional Gaussian model corresponds to at least one second video, and the difference between each third three-dimensional Gaussian model and the second video corresponding to each third three-dimensional Gaussian model satisfies the difference condition.

[0124] Among them, the acquisition unit 801, the adjustment unit 802 and the storage unit 803 can all be implemented by software, or can be implemented by hardware. Exemplarily, the following takes the acquisition unit 801 as an example to introduce the implementation of the acquisition unit 801. Similarly, the implementation of the adjustment unit 802 and the storage unit 803 can refer to the implementation of the acquisition unit 801.

[0125] As an example of a software functional unit, the acquisition unit 801 may include code running on a computing instance. The computing instance may include at least one of a physical host (computing device), a virtual machine, and a container. Furthermore, the computing instance may be one or more. For example, the acquisition unit 801 may include code running on multiple hosts / virtual machines / containers. The multiple hosts / virtual machines / containers used to run the code may be distributed in the same region or in different regions. Furthermore, the multiple hosts / virtual machines / containers used to run the code may be distributed in the same availability zone (AZ) or in different AZs, each AZ including one data center or multiple data centers with close geographical locations. Typically, a region may include multiple AZs.

[0126] Similarly, multiple hosts / virtual machines / containers used to run the code can be distributed in the same virtual private cloud (VPC) or in multiple VPCs. Usually, a VPC is set up in a region. For cross-region communication between two VPCs in the same region and between VPCs in different regions, a communication gateway needs to be set up in each VPC to achieve interconnection between VPCs through the communication gateway.

[0127] As an example of a hardware functional unit, the acquisition unit 801 may include at least one computing device, such as a server, etc. Alternatively, the acquisition unit 801 may also be a device implemented using an application-specific integrated circuit (ASIC) or a programmable logic device (PLD). The PLD may be a complex programmable logical device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof.

[0128] The multiple computing devices included in the acquisition unit 801 can be distributed in the same region or in different regions. The multiple computing devices included in the acquisition unit 801 can be distributed in the same AZ or in different AZs. Similarly, the multiple computing devices included in the acquisition unit 801 can be distributed in the same VPC or in multiple VPCs. The multiple computing devices can be any combination of computing devices such as servers, ASICs, PLDs, CPLDs, FPGAs, and GALs.

[0129] It should be noted that, in other embodiments, the acquisition unit 801 can be used to execute any step in the method for processing a video, the adjustment unit 802 can be used to execute any step in the method for processing a video, and the storage unit 803 can be used to execute any step in the method for processing a video. The steps that the acquisition unit 801, the adjustment unit 802, and the storage unit 803 are responsible for implementing can be specified as needed, and the acquisition unit 801, the adjustment unit 802, and the storage unit 803 respectively implement different steps in the method for processing a video to implement all functions of the apparatus 800 for processing a video.

[0130] In the embodiment of the present application, since the adjustment unit adjusts the appearance styles of the multiple frames of first object images included in the first video, at least one second video corresponding to at least one appearance style is obtained, so the acquisition unit acquires at least one first three-dimensional Gaussian model with different appearance styles based on the original video and at least one second video. In other words, one or more first three-dimensional Gaussian models with different appearance styles can be constructed, thereby enriching the appearance and style of the three-dimensional Gaussian model.

[0131] See also Fig. 9 , the present application embodiment provides a computing device 900. For example, the computing device 900 may be Figure 5 The computing device shown, or Figure 6 A computing device that performs method 600, or Figure 7 A computing device is shown performing method 700 .

[0132] like Fig. 9 As shown, the computing device 900 includes: a bus 902, a processor 904, a memory 906, and a communication interface 908. The processor 904, the memory 906, and the communication interface 908 communicate through the bus 902. The computing device 900 can be a server or a terminal device. It should be understood that the present application does not limit the number of processors and memories in the computing device 900.

[0133] The bus 902 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Fig. 9 The bus 902 is represented by only one line, but it does not mean that there is only one bus or one type of bus. The bus 902 may include a path for transmitting information between various components of the computing device 900 (for example, the processor 904, the memory 906, and the communication interface 908).

[0134] The processor 904 may include any one or more of a central processing unit (CPU), a graphics processing unit (GPU), a microprocessor (MP), or a digital signal processor (DSP).

[0135] The memory 906 may include a volatile memory, such as a random access memory (RAM). The memory 906 may also include a non-volatile memory, such as a read-only memory (ROM), a flash memory, a hard disk drive (HDD), or a solid state drive (SSD).

[0136] See also Fig. 9 , the memory 906 stores executable program codes, and the processor 904 executes the executable program codes to respectively implement Figure 8 The functions of the acquisition unit 801, the adjustment unit 802 and the storage unit 803 in the device 800 shown in the figure are implemented to realize the method provided by any of the above embodiments. That is, the memory 906 stores instructions for executing the method provided by any of the above embodiments. Or, The communication interface 908 uses a transceiver module such as, but not limited to, a network interface card or a transceiver to implement communication between the computing device 900 and other devices or communication networks.

[0137] The embodiment of the present application also provides a computing device cluster. The computing device cluster includes at least one computing device. The computing device can be a server, such as a central server, an edge server, or a local server in a local data center. In some embodiments, the computing device can also be a terminal device such as a desktop computer, a laptop computer, or a smart phone.

[0138] like Fig.10 As shown, the computing device cluster includes at least one computing device 900. The memory 906 in one or more computing devices 900 in the computing device cluster may store the same instructions for executing the method for processing a video provided in any of the above embodiments.

[0139] In some possible implementations, the memory 906 of one or more computing devices 900 in the computing device cluster may also store partial instructions for executing the above-mentioned method for processing video. In other words, the combination of one or more computing devices 900 can jointly execute instructions for executing the method provided in any of the above-mentioned embodiments.

[0140] The memory 906 in different computing devices 900 in the computing device cluster may store different instructions for executing the following respectively: Figure 8 Part of the functions of the apparatus 800 for processing video shown. That is, the instructions stored in the memory 906 in different computing devices 900 can implement the functions of one or more units among the obtaining unit 801, the adjusting unit 802 and the saving unit 803.

[0141] In some possible implementations, one or more computing devices in the computing device cluster may be connected via a network, which may be a wide area network or a local area network. Fig.11 A possible implementation is shown. Fig.11 As shown, two computing devices 900A and 900B are connected via a network. Specifically, they are connected to the network via a communication interface in each computing device.

[0142] In this type of possible implementation, the memory 906 in the computing device 900A stores the execution Figure 8 The instructions for the functions of the acquisition unit 801 and the adjustment unit 802 in the embodiment shown in the figure are shown in the figure. Figure 8 Instructions for the functionality of the storage unit 803 in the illustrated embodiment.

[0143] Fig.11The connection method between the computing device clusters shown may be based on the needs of the method for processing videos provided in the present application (eg, storing and / or sending and receiving large amounts of data), and therefore, it is considered that the functions implemented by the storage unit 803 may be executed by the computing device 900B.

[0144] It should be understood that Fig.11 The functions of the computing device 900A shown in FIG. 9A may also be completed by multiple computing devices 900. Similarly, the functions of the computing device 900B may also be completed by multiple computing devices 900.

[0145] The present application embodiment also provides another computing device cluster. The connection relationship between the computing devices in the computing device cluster can be similar to that of Fig.11 The connection mode of the computing device cluster is different in that the memory 906 in one or more computing devices 900 in the computing device cluster may store the same instructions for executing the method for processing a video provided in any of the above embodiments.

[0146] In some possible implementations, the memory 906 of one or more computing devices 900 in the computing device cluster may also store partial instructions for executing the method provided in any of the above embodiments. In other words, the combination of one or more computing devices 900 may jointly execute instructions for executing the method provided in any of the above embodiments.

[0147] The present application also provides a computer program product including instructions. The computer program product may be software or a program product including instructions that can be run on a computing device or stored in any available medium. When the computer program product is run on at least one computing device, the at least one computing device executes the method provided in any of the above embodiments.

[0148] The embodiment of the present application also provides a computer-readable storage medium. The computer-readable storage medium can be any available medium that can be stored by a computing device or a data storage device such as a data center containing one or more available media. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state hard disk). The computer-readable storage medium includes instructions that instruct the computing device to execute the method provided in any of the above embodiments.

[0149] A person skilled in the art will understand that all or part of the steps to implement the above embodiments may be accomplished by hardware or by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.

[0150] The above description is only an optional embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for processing a video, characterized in that: The method comprises: Acquire a first video based on an original video, wherein the original video includes multiple frames of pictures obtained by shooting a target object from multiple perspectives, and the first video includes multiple frames of first object images, and the multiple frames of first object images are images of the target object in the multiple frames of pictures; Adjusting the appearance styles of the multiple frames of first object images included in the first video to obtain at least one second video corresponding to at least one appearance style, each second video includes multiple frames of second object images, and the appearance styles of the multiple frames of second object images included in each second video are respectively the appearance styles corresponding to each second video; At least one first three-dimensional Gaussian model is acquired based on the original video and the at least one second video, wherein the appearance style of the at least one first three-dimensional Gaussian model is the same as the at least one appearance style.

2. The method according to claim 1, characterized in that The at least one appearance style comprises a target appearance style; The step of adjusting the appearance styles of the plurality of frames of first object images included in the first video to obtain at least one second video corresponding to at least one appearance style includes: Based on a target filter template, adjusting a parameter value of at least one illumination parameter of each frame of the first object image included in the first video to obtain a second video corresponding to the target appearance style, wherein the target filter template is used to store a standard parameter value of the at least one illumination parameter, and the standard parameter value is used to indicate the target appearance style; or Based on the image style modification model and the first video, the appearance styles of the multiple frames of first object images included in the first video are modified to the target appearance style, so as to obtain a second video corresponding to the target appearance style.

3. The method according to claim 1 or 2, characterized in that The at least one appearance style includes a target appearance style, and the second video corresponding to the target appearance style includes a plurality of frames of second object images corresponding to the plurality of frames of pictures; The acquiring at least one first three-dimensional Gaussian model based on the original video and the at least one second video includes: Obtaining the camera position and camera parameters corresponding to each frame of the original video; Based on the camera pose, camera parameters and the second object image corresponding to each frame of the image, a first three-dimensional Gaussian model of the target appearance style is constructed.

4. The method according to claim 1 or 2, characterized in that: The at least one appearance style comprises a target appearance style; The acquiring at least one first three-dimensional Gaussian model based on the original video and the at least one second video includes: Acquire a second three-dimensional Gaussian model of the target object based on the original video; Adjusting a parameter value of at least one illumination parameter of the second three-dimensional Gaussian model to obtain a third three-dimensional Gaussian model; When a difference between the third 3D Gaussian model and the second video corresponding to the target appearance style satisfies a difference condition, the second 3D Gaussian model and the third 3D Gaussian model are superimposed to obtain a first 3D Gaussian model of the target appearance style.

5. The method according to claim 4, characterized in that The method further comprises: When the difference between the third three-dimensional Gaussian model and the second video corresponding to the target appearance style does not satisfy the difference condition, adjusting a parameter value of at least one illumination parameter of the third three-dimensional Gaussian model to obtain a fourth three-dimensional Gaussian model; When the difference between the fourth 3D Gaussian model and the second video corresponding to the target appearance style satisfies the difference condition, the second 3D Gaussian model and the fourth 3D Gaussian model are superimposed to obtain a first 3D Gaussian model of the target appearance style.

6. The method according to claim 4, characterized in that The at least one illumination parameter of the second three-dimensional Gaussian model includes one or more of the following: color or transparency.

7. A device for processing video, characterized in that: The device comprises: an acquisition unit, configured to acquire a first video based on an original video, wherein the original video includes a plurality of frames of pictures obtained by photographing a target object from a plurality of perspectives, and the first video includes a plurality of frames of first object images, wherein the plurality of frames of first object images are images of the target object in the plurality of frames of pictures; an adjusting unit, configured to adjust the appearance styles of the multiple frames of first object images included in the first video to obtain at least one second video corresponding to at least one appearance style, each second video including multiple frames of second object images, and the appearance styles of the multiple frames of second object images included in each second video are respectively the appearance styles corresponding to each second video; The acquisition unit is further used to acquire at least one first three-dimensional Gaussian model based on the original video and the at least one second video, and the appearance style of the at least one first three-dimensional Gaussian model is the same as the at least one appearance style.

8. A computing device cluster, characterized in that: comprising at least one computing device, each computing device comprising a processor and a memory; The processor of the at least one computing device is configured to execute instructions stored in the memory of the at least one computing device, so that the computing device cluster executes the method according to any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that: The method comprises computer program instructions. When the computer program instructions are executed by a computing device cluster, the computing device cluster performs the method according to any one of claims 1 to 6.

10. A computer program product comprising instructions, characterized in that When the instruction is executed by the computing device cluster, the computing device cluster executes the method according to any one of claims 1 to 6.

Images