Image composition device, image composition method, and program

Abstract

To provide an image processing technique that reduces sense of incongruity in a composite image of a foreground image and a background image.SOLUTION: One aspect of the present disclosure relates to an image composition device, which includes: an image acquisition unit that acquires a foreground image and a background image; a composite image generation unit that corrects a composite image of the foreground image and the background image and generates a corrected composite image; and an image output unit that outputs the corrected composite image.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present disclosure relates to an image synthesizing apparatus, an image synthesizing method, and a program.

Background Art

[0002] With the improvement of information technology, image processing technology for processing image data has evolved rapidly. Users can perform various data processing on image data by using image editing tools or the like on user terminals such as smartphones, tablets, and personal computers. For example, a user can operate an image editing tool to adjust the contrast or brightness of an image, or extract a person or the like captured in the image.

[0003] In addition, with the evolution of recent machine learning technologies, research and development of various machine learning models for image processing have been promoted, and machine learning models have been used for image processing.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Patent Document 2

Patent Document 3

Summary of the Invention

Problems to be Solved by the Invention

[0005] For example, when the user replaces the background image captured in the image of the subject with another background image desired by the user, an image processing for synthesizing the foreground image of the captured subject and the background image to be replaced is performed. However, due to differences in color tone between the foreground image and the background image, inconsistency of light sources, unnatural focus, difference in angle of view, etc., there may be a sense of discomfort in the generated composite image. There is a need for an image processing technique for alleviating the sense of discomfort in the composite image of such a foreground image and a background image.

[0006] An object of the present disclosure is to provide an image processing technique for alleviating the sense of discomfort in a composite image of a foreground image and a background image.

Means for Solving the Problems

[0007] One aspect of the present disclosure relates to an image synthesizing apparatus including an image acquisition unit that acquires a foreground image and a background image, a composite image generation unit that corrects a composite image of the foreground image and the background image and generates a corrected composite image, and an image output unit that outputs the corrected composite image.

Effects of the Invention

[0008] According to the present disclosure, it is possible to provide an image processing technique for alleviating the sense of discomfort in a composite image of a foreground image and a background image.

Brief Description of the Drawings

[0009]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

[0010] Hereinafter, embodiments of the present disclosure will be described with reference to the drawings.

[0011] In the following embodiments, an image composition device that generates a composite image in which the sense of incongruity caused by differences in shooting conditions and the like between a foreground image and a background image is alleviated is disclosed.

[0012] [Overview of the Present Disclosure] In the embodiments described below, as shown in FIG. 1, the image synthesizing apparatus 100 synthesizes a foreground image 10 in which a subject to be synthesized such as a person is imaged and a background image 20 in which the background of the subject to be synthesized is imaged, and corrects the synthesized image so as to alleviate the discomfort of the synthesized image that may occur due to differences in shooting conditions such as the foreground image 10 and the background image 20, and generates a synthesized image 30. For example, in the illustrated example, a blur process is applied to the background image 20 so as to adjust the sense of distance between the subject of the foreground image 10 and the building of the background image 20, and the subject of the foreground image 10 is superimposed on the background image 20 to which the blur process has been applied.

[0013] In the following embodiments, the image synthesizing apparatus 100 generates a synthesized image subjected to correction processes such as angle-of-view adjustment, blur processing, color correction, and light source adjustment in order to alleviate the discomfort in the synthesized image that may occur due to differences in shooting conditions such as differences in color tone between the foreground image 10 and the background image 20, inconsistency of light sources, unnatural focus, and differences in angle of view. The synthesized image corrected in this way can be generated, for example, by deleting the background in which the interior of the user's home etc. is shown from a self-shot image taken by the user at home etc. and synthesizing it with the user's favorite background image. For example, it becomes possible to easily create a profile image etc. using the image synthesizing apparatus 100.

[0014] Here, the image synthesizing apparatus 100 may be realized by a computing device such as a server, a personal computer (PC), a smartphone, or a tablet, and may have, for example, a hardware configuration as shown in FIG. 2. For example, the image synthesizing apparatus 100 may be mounted on a user terminal or may be realized as a server on the cloud. That is, the image synthesizing apparatus 100 includes a drive device 101, a storage device 102, a memory device 103, a processor 104, a user interface (UI) device 105, and a communication device 106 that are interconnected via a bus B.

[0015] Programs or instructions for realizing various functions and processes in the image synthesizing apparatus 100 may be stored in a removable storage medium such as a CD-ROM (Compact Disk-Read Only Memory) or a flash memory. When the storage medium is set in the drive device 101, the program or instruction is installed from the storage medium via the drive device 101 into the storage device 102 or the memory device 103. However, the program or instruction does not necessarily have to be installed from the storage medium and may be downloaded from any external device via a network or the like.

[0016] The storage device 102 is realized by a hard disk drive or the like and stores files, data, etc. used for execution of the program or instruction together with the installed program or instruction.

[0017] The memory device 103 is realized by a random access memory, a static memory, or the like, and when the program or instruction is activated, reads and stores the program or instruction, data, etc. from the storage device 102. The storage device 102, the memory device 103, and the removable storage medium may be collectively referred to as a non-transitory storage medium.

[0018] The processor 104 may be realized by one or more CPUs (Central Processing Unit), GPUs (Graphics Processing Unit), processing circuitry, etc. that may be composed of one or more processor cores, and executes various functions and processes of the image synthesizing apparatus 100 according to data such as programs, instructions, and parameters necessary for executing the program or instruction stored in the memory device 103.

[0019] The user interface (UI) device 105 may be composed of input devices such as a keyboard, a mouse, a camera, a microphone, etc., output devices such as a display, a speaker, a headset, a printer, etc., and input / output devices such as a touch panel, and realizes an interface between the user and the image synthesizing device 100. For example, the user may operate the image synthesizing device 100 by operating a keyboard, a mouse, etc. on the GUI (Graphical User Interface) displayed on the display or the touch panel.

[0020] The communication device 106 is realized by various communication circuits that execute wired and / or wireless communication processing with communication networks such as external devices, the Internet, a LAN (Local Area Network), and a cellular network.

[0021] However, the above-described hardware configuration is merely an example, and the image synthesizing device 100 according to the present disclosure may be realized by any other appropriate hardware configuration.

[0022] [Image Synthesizing Device] Next, the image synthesizing device 100 according to an embodiment of the present disclosure will be described. FIG. 3 is a block diagram showing the functional configuration of the image synthesizing device 100 according to an embodiment of the present disclosure. As shown in FIG. 3, the image synthesizing device 100 includes an image acquisition unit 110, a composite image generation unit 120, and an image output unit 130. Each functional unit of the image acquisition unit 110, the composite image generation unit 120, and the image output unit 130 may be realized by a computer program stored in the memory device 103 of the image synthesizing device 100 being executed by the processor 104.

[0023] The image acquisition unit 110 acquires the foreground image 10 and the background image 20. For example, when the image synthesis device 100 is mounted on a user terminal, the image acquisition unit 110 may acquire, as the foreground image 10 and / or the background image 20, an image selected from an image group stored in, for example, the image folder of the user terminal. Alternatively, when the image synthesis device 100 is realized as a server on the cloud, the image acquisition unit 110 may acquire, as the foreground image 10 and / or the background image 20, an image selected by the user from a user terminal, an image database, or the like.

[0024] Specifically, the image acquisition unit 110 acquires, from a user or the like of the image synthesis device 100, an image 11 of a subject to be synthesized and a background image 20 of the background to be synthesized. Typically, the image 11 may include not only the subject but also the background around the subject at the time of shooting. In order to delete the background around the subject from the image 11 and extract only the subject, the image acquisition unit 110 extracts the subject to be synthesized from the acquired image 11 and acquires a foreground image 10 composed of the extracted subject.

[0025] For example, when an image 11 of a subject as shown in FIG. 4 is acquired from a user terminal, the image acquisition unit 110 inputs the image 11 to a mask image generation model 61 trained to extract a subject such as a person from the input image and generate a mask image obtained by masking the extracted subject, and acquires a mask image 12 of the subject from the mask image generation model 61. Then, the image acquisition unit 110 can synthesize the mask image 12 and the image 11 to generate the foreground image 10. Such a mask image generation model 61 can be realized, for example, as any machine learning model such as a neural network. The image acquisition unit 110 passes the generated foreground image 10 and the background image 20 to the composite image generation unit 120. Note that the generation of the foreground image 10 of the subject from the image 11 is not limited to this, and any known extraction technique may be used.

[0026] The composite image generation unit 120 corrects the composite image of the foreground image and the background image, and generates a corrected composite image. Specifically, the composite image generation unit 120 may perform image processing such as angle adjustment, blurring processing, color correction, and light source adjustment on the foreground image 10, the background image 20, and / or the composite image of the foreground image 10 and the background image 20.

[0027] First, regarding the angle adjustment, in one embodiment, the composite image generation unit 120 may perform angle adjustment on the background image 20. Specifically, as shown in FIG. 5, the composite image generation unit 120 uses an angle adjustment model 62 trained to generate an angle-adjusted image 21 in which the angle and / or position of the object imaged in the input image is adjusted, and performs angle adjustment on the background image 20. In the foreground image 10, the subject typically faces the front. Therefore, the angle adjustment may also adjust the angle and / or position of the object in the background image 20 so that the object imaged in the background image 20 faces the front.

[0028] Here, the angle adjustment model 62 may be realized as any machine learning model such as a neural network, and specifically, may be realized by a model disclosed in "User-Controllable Latent Transformer for StyleGAN Image Layout Editing" (https: / / github.com / endo-yuki-t / UserControllableLT) or the like.

[0029] Note that the angle-of-view adjustment is not limited to the use of the angle-of-view adjustment model 62. For example, the composite image generation unit 120 may acquire a wide-angle background image 20 such as a 360-degree image, and extract, as an angle-of-view adjusted background image 22, an image area selected by the user from the wide-angle background image 20. For example, in the example shown in FIG. 6, the background image 20 is a 360-degree wide-angle image, and the composite image generation unit 120 allows the user to select background areas 22_1, 22_2, 22_3, etc. with an angle of view suitable for the foreground image 10 from this background image 20, and may select the selected background area 22_1, 22_2, or 22_3 as the angle-of-view adjusted background image 22. Thereby, it becomes possible to perform angle-of-view adjustment according to the user's preference.

[0030] Next, regarding the blurring process, in one embodiment, the composite image generation unit 120 may estimate the depth of field of the background image 20 and perform a blurring process on the background image 20 with different blurring intensities according to the depth of field. Specifically, the composite image generation unit 120 uses a depth estimation model 63 trained to estimate the depth of each object imaged in the input image and generate a depth estimation image indicating the estimated depth, and performs a blurring process with different intensities according to the depth.

[0031] For example, as shown in FIG. 7, the composite image generation unit 120 inputs the background image 20 to the depth estimation model 63 and acquires a depth estimation image 23 as a mask image from the depth estimation model 63. In the illustrated depth estimation image 23, each object imaged in the background image 20 is grayscaled according to the depth, and the farther the object is from the imaging position, i.e., the greater the depth, the darker it is colored, and the closer the object is to the imaging position, i.e., the smaller the depth, the lighter it is colored. Then, the composite image generation unit 120 synthesizes the acquired depth estimation image 23 and the background image 20 to generate a blurred image 24. In the blurring process, blurring is applied so that the greater the depth of the object in the background image 20, the greater the blurring intensity, and the smaller the depth of the object in the background image 20, the smaller the blurring intensity.

[0032] Note that the blurring process here is not limited to the use of the depth estimation image 23. For example, it is also possible to blur the background (smoothing) by applying a function of OpenCV to the background image 20. The blurring intensity can be adjusted by adjusting the function to be applied and each parameter.

[0033] Next, regarding color correction, in one embodiment, the composite image generation unit 120 may perform color correction on the foreground image 10 so as to match the color tone of the background image 20. Specifically, the composite image generation unit 120 uses a color correction model 64 trained to harmonize the color tone according to the background in the input image and generate a color-corrected image, and performs color correction on the composite image of the foreground image 10 and the background image 20. For example, as shown in FIG. 8, the composite image generation unit 120 inputs the (blurred-processed) composite image 25 of the foreground image 10 and the background image 20 to the color correction model 64, and may obtain a color-corrected composite image 26 from the color correction model 64 such that the color tone of the subject of the foreground image 10 matches the color tone of the background image 20.

[0034] Incidentally, although the color correction model 64 described above accepts the composite image 25 of the foreground image 10 and the background image 20 as input, it is not limited thereto. For example, as shown in FIG. 9, a color correction model 65 that accepts two images, the foreground image 10 and the background image 20, separately as input and outputs a color-corrected composite image 26 may be used. Here, the color correction model 65 may be realized as any machine learning model such as a neural network trained to generate a composite image in which the subject of the foreground image is color-corrected to match the color tone of the background image for the input foreground image and background image. Such a color correction model 65 may be realized as any machine learning model such as a neural network, and specifically, may be realized by a model disclosed in "Semi-supervised Parametric Real-world Image Harmonization" (https: / / github.com / adobe / PIH) or the like. Thereby, the color tone of the subject can be matched to the color tone of the background image, and the sense of incongruity between the subject and the background in the composite image 26 can be alleviated.

[0035] Also, in one embodiment, the composite image generation unit 120 may add additional information to the composite image 25 and perform color correction on the composite image 25 based on the additional information. Here, the additional information is any information related to color correction, and for example, it may be a color frame added to the composite image 25. The composite image generation unit 120 inputs the composite image 25 with the color frame added thereto into a color correction model 66 trained to generate an image color-corrected according to the color frame from the image with the color frame added, and may obtain a color-corrected composite image 26. Here, the color correction model 66 can accept not only an image with a color frame added but also an image without a color frame added. For an image with a color frame added, it outputs an image color-corrected according to the thickness and color of the color frame, and for an image without a color frame added, it may output an image color-corrected in the same manner as the above-described color correction model 64. For example, the color correction model 66 may be implemented as any machine learning model such as a neural network, and specifically, it may be trained to adjust the color tone of the subject in the foreground image 10 according to the color tone of the color frame or adjust the brightness of the subject in the foreground image 10 according to the thickness of the color frame.

[0036] For example, as shown in FIG. 10, when the composite image generation unit 120 obtains the composite image 25 with the color frame 27 added, it may input the obtained composite image 25 with the color frame 27 added into the color correction model 66 and obtain a composite image 26 color-corrected according to the color frame 27 from the color correction model 66. Alternatively, as shown in FIG. 11, when the composite image generation unit 120 obtains the foreground image 10 and the background image 20 with the color frame 27 added, it may input the obtained foreground image 10 and the background image 20 with the color frame 27 added into the color correction model 66 and obtain a composite image 26 color-corrected according to the color frame 27 from the color correction model 66. Here, although the color frame 27 is added to the background image 20, the present disclosure is not limited thereto, and the color frame 27 may be added to the foreground image 10. Note that such additional information may be specified by the user or may be automatically generated based on the foreground image 10 and / or the background image 20. For example, the additional information may be generated based on the difference in brightness between the foreground image 10 and the background image 20.

[0037] Alternatively, the composite image generation unit 120 may calculate the average pixel value of the background image 20, and synthesize the calculated average pixel value and the foreground image 10 to generate a color-corrected composite image 26. For example, as shown in FIG. 12, the composite image generation unit 120 may generate an average image 27 composed of the average pixel values of the background image 20, and synthesize the generated average image 27 and the foreground image 10. Specifically, the composite image generation unit 120 may add α% (0 < α ≤ 100) of the average pixel value to the subject of the foreground image 10.

[0038] Note that the color correction according to the present disclosure is not limited to the color correction process described above. For example, rule-based color correction may be used, and in this case, images may be superimposed in a blend mode. In such a blend mode function, a color layer extracted from the background image 20 is superimposed on the foreground image 10, and the color tone and brightness of the foreground image 10 can be changed by selecting a synthesis method. The color tone after processing can be changed according to the type of blend mode and / or the transparency of the color layer.

[0039] Next, regarding the light source adjustment, in one embodiment, the composite image generation unit 120 may perform light source adjustment on the composite image 28 so as to match the light source positions of the foreground image 10 and the background image 20. Typically, the light source positions of the foreground image 10 and the background image 20 do not necessarily match. When the foreground image 10 and the background image 20 captured from different light source positions are synthesized without light source adjustment, a sense of incongruity may occur in the composite image 28. Therefore, the composite image generation unit 120 may perform light source adjustment on the composite image 28 so as to match the light source positions of the foreground image 10 and the background image 20, and generate a light source-adjusted composite image 29.

[0040] For example, as shown in FIG. 13, the composite image generation unit 120 may perform light source adjustment on the composite image 28 of the foreground image 10 and the background image 20 by using the light source adjustment model 67 trained to receive, as an input, a composite image of two images with different light source positions and generate a composite image with the light source positions matched. Specifically, the composite image generation unit 120 inputs the composite image 28 of the foreground image 10 and the background image 20 captured from different light source positions into the light source adjustment model 67, and obtains the light source-adjusted composite image 29 with the light source positions matched from the light source adjustment model 67. For example, such a light source adjustment model 67 may be realized as any machine learning model such as a neural network, and specifically, may be realized by the model disclosed in "Face Relighting with Geometrically Consistent Shadow" (https: / / github.com / andrewhou1 / GeomConsistentFR).

[0041] Note that the individual image processes of the above-described angle-of-view adjustment, blurring process, color correction, and light source adjustment do not necessarily need to be executed in the above-described order, and may be executed in any appropriate order. Also, the images input to the angle-of-view adjustment model 62, the depth estimation model 63, the color correction models 64 to 66, and the light source adjustment model 67 may be the images processed by the immediately preceding image process according to the execution order of the individual image processes. For example, when color correction is executed after the execution of light source adjustment, the images input to the color correction models 64 to 66 are the light source-adjusted images.

[0042] In the above-described embodiments, the composite image generation unit 120 performs individual image processing such as angle-of-view adjustment, blurring processing, color correction, and light source adjustment on the foreground image 10, the background image 20, and / or the composite image of the foreground image 10 and the background image 20, and obtains the composite image 30 that is corrected to reduce the sense of incongruity. However, the present disclosure is not limited thereto. For example, the composite image generation unit 120 may input the foreground image 10 and the background image 20 to an image correction model 70 trained to correct the composite image of the foreground image and the background image, and obtain the corrected composite image 30. That is, the composite image generation unit 120 receives the foreground image 10 and the background image 20 as inputs without individually performing image processing such as angle-of-view adjustment, blurring processing, color correction, and light source adjustment, and uses an image correction model 70 that outputs a composite image 30 corrected to reduce the sense of incongruity in the composite image that may occur due to differences in shooting conditions such as differences in color tone between the foreground image 10 and the background image 20, inconsistencies in light sources, unnatural focus, and differences in angle-of-view, to obtain the corrected composite image 30 from the foreground image 10 and the background image 20.

[0043] For example, the image correction model 70 may be realized as a machine learning model trained according to the error between the output result of the foreground image and the background image of the training data set and the corrected composite image of the training data set, using a training data set consisting of a foreground image, a background image, and a corrected composite image. For example, when the image correction model 70 is realized as a neural network model, the parameters of the image correction model 70 can be adjusted according to the error backpropagation method according to the error between the output result of the foreground image and the background image of the training data set input to the image correction model 70 and the corresponding corrected composite image of the training data set.

[0044] According to such an image correction model 70, as shown in FIG. 14, it becomes possible to obtain a composite image 30 corrected to reduce the sense of incongruity due to composition from the foreground image 10 and the background image 20 without applying individual image processing to the foreground image 10 and the background image 20.

[0045] The image output unit 130 outputs the corrected composite image 30. Specifically, the image output unit 130 displays, for example, on a display, the composite image 30 corrected so that the sense of incongruity generated by the composite image generation unit 120 is alleviated, and presents the corrected composite image 30 to the user.

[0046] Here, the image output unit 130 may display both the corrected composite image generated by individual image processing such as angle of view adjustment, blurring process, color correction, light source adjustment, etc. and the corrected composite image generated by the image correction model 70 on the display so that it can be selected by the user. That is, as shown in FIG. 15, the image output unit 130 corrects the corrected composite image 30_1 obtained by performing image processing on the foreground image 10, the background image 20, or the composite image, and the foreground image 10 and the background image 20 by the image correction model 70 trained to correct the composite image of the foreground image and the background image. The corrected composite image 30_2 obtained from may be displayed so that the user can select it.

[0047] Specifically, when the composite image generation unit 120 acquires the foreground image 10 and the background image 20, in parallel with performing one or more individual image processes such as the above-described angle of view adjustment, blurring process, color correction, light source adjustment, etc. on the acquired foreground image 10 and background image 20, the acquired foreground image 10 and background image 20 are input to the image correction model 70 and the generation process by the image correction model 70 is executed. The image output unit 130 may display the corrected composite image 30_1 generated by individual image processing and the corrected composite image 30_2 generated by the image correction model 70 on the display together with check boxes so that the user can select them. Thereby, the user can select a composite image that suits his / her preference.

[0048] [Image Composition Processing] Next, an image synthesis process according to an embodiment of the present disclosure will be described. FIG. 16 is a flowchart showing the image synthesis process according to an embodiment of the present disclosure. The image synthesis process may be executed, for example, by an image synthesis apparatus 100, and more specifically, may be realized by a processor 104 of the image synthesis apparatus 100 executing a computer program or instruction stored in a memory apparatus 103.

[0049] In step S101, the image synthesis apparatus 100 acquires a foreground image 10 and a background image 20. Specifically, the image synthesis apparatus 100 acquires from the user a foreground image 10 in which a subject to be synthesized is imaged and a background image 20 in which a background of the synthesis target is imaged. For example, when the image synthesis apparatus 100 is mounted on a user terminal, or when a program, software, application, etc. that realizes the function of the image synthesis apparatus 100 is installed on the user terminal, the foreground image 10 and the background image 20 may be selected by the user from an image folder in the user terminal.

[0050] Alternatively, when the image synthesis apparatus 100 is realized as a server on the cloud, the image synthesis apparatus 100 may acquire the foreground image 10 and / or the background image 20 from, for example, a user terminal and / or an image database.

[0051] In step S102, the image synthesis apparatus 100 corrects a composite image of the foreground image 10 and the background image 20 to generate a corrected composite image 30. Specifically, the image synthesis apparatus 100 performs one or more individual image processes such as angle of view adjustment, blurring process, color correction, light source adjustment, etc. on the foreground image 10, the background image 20, and / or their composite images, and corrects the composite image in which the foreground image 10 and the background image 20 are simply superimposed to generate a corrected composite image 30 that alleviates the sense of incongruity. Also, instead of or in parallel with such individual image processes, the image synthesis apparatus 100 may use an image correction model 70 trained using a training data set composed of the foreground image, the background image, and the composite image with the sense of incongruity alleviated to generate a corrected composite image 30 that alleviates the sense of incongruity.

[0052] In step S103, the image synthesizing apparatus 100 outputs the corrected synthesized image 30. Specifically, the image synthesizing apparatus 100 displays the synthesized image 30_1 corrected by individual image processing to alleviate the sense of incongruity and / or the synthesized image 30_2 generated from the image correction model 70 on the display of the user terminal, and / or stores them in the image folder of the user terminal. Further, the synthesized image 30_1 generated by individual image processing and the synthesized image 30_2 generated from the image correction model 70 may be displayed so as to be selectable by the user.

[0053] According to the above-described embodiment, in order to alleviate the sense of incongruity in the synthesized image that may occur due to differences in shooting conditions such as differences in color tone between the foreground image and the background image, inconsistencies in light sources, unnatural focus, and differences in shooting angles, it is possible to generate a synthesized image subjected to correction processing such as shooting angle adjustment, blurring processing, color correction, and light source adjustment.

[0054] As described above, the embodiments of the present disclosure have been described in detail. However, the present disclosure is not limited to the specific embodiments described above, and various modifications and changes are possible within the scope of the gist of the present disclosure described in the claims.

Explanation of Reference Numerals

[0055] 100 Image synthesizing apparatus 110 Image acquisition unit 120 Synthesized image generation unit 130 Image output unit

Claims

1. an image acquisition unit that acquires a foreground image and a background image; a composite image generating unit that corrects a composite image of the foreground image and the background image and generates a corrected composite image; an image output unit that outputs the corrected composite image; An image synthesis device comprising:

2. The image synthesis device according to claim 1 , wherein the synthetic image generation unit performs one or more of image processing including an angle of view adjustment, a blurring process, a color correction, and a light source adjustment on the foreground image, the background image, or the synthetic image.

3. The image synthesis device according to claim 2 , wherein the synthetic image generation unit performs an angle of view adjustment on the background image.

4. The image synthesis device according to claim 2 , wherein the synthetic image generation unit estimates a depth of field of the background image, and performs blurring on the background image with a blurring strength that varies depending on the depth of field.

5. The image synthesis device according to claim 2 , wherein the synthetic image generating unit performs color correction on the foreground image so as to match a color tone of the background image.

6. The image synthesis device according to claim 2 , wherein the synthetic image generating section adds additional information to the synthetic image, and performs color correction on the synthetic image based on the additional information.

7. the additional information is a color frame to be added to the composite image, 7. The image synthesis device according to claim 6, wherein the composite image generation unit inputs the composite image to which the color frame has been added to a color correction model that has been trained to generate an image that has been color-corrected in accordance with the color frame from an image to which the color frame has been added, and obtains a color-corrected composite image.

8. The image synthesis device according to claim 2 , wherein the synthetic image generation unit performs light source adjustment on the synthetic image so that light source positions of the foreground image and the background image coincide with each other.

9. 2. The image synthesis device according to claim 1, wherein the composite image generation unit inputs the foreground image and the background image into an image correction model trained to correct a composite image of the foreground image and the background image, and obtains the corrected composite image.

10. 2. The image synthesis device according to claim 1, wherein the image output unit displays to a user in a selectable manner a first corrected composite image obtained by performing image processing on the foreground image, the background image, or the composite image, and a second corrected composite image obtained from the foreground image and the background image using an image correction model trained to correct the composite image of the foreground image and the background image.

11. Obtaining a foreground image and a background image; correcting a composite image of the foreground image and the background image to generate a corrected composite image; outputting the corrected composite image; and A computer implemented image synthesis method.

12. Obtaining a foreground image and a background image; correcting a composite image of the foreground image and the background image to generate a corrected composite image; outputting the corrected composite image; and A program that causes a computer to execute the following.

Images