JP7143419B2 - Image processing device, image processing method, and program - Google Patents

Description

The present invention relates to an image processing device, an image processing method, and a program, and more particularly to an image processing device, an image processing method, and a program for generating three-dimensional image data from a two-dimensional image.

Conventionally, three-dimensional image data is generated by using the main subject of an input two-dimensional image as the foreground and adding depth information to the foreground and background.

Here, in order to generate a large amount of three-dimensional image data, it is necessary to efficiently generate three-dimensional image data by automatic image processing. On the other hand, if the boundary between the foreground and the background is sharply cut out, the boundary becomes conspicuous and the generated three-dimensional image data has a strange feeling.

Various techniques have been proposed for the purpose of suppressing the sense of incongruity of the three-dimensional image data.

For example, the technique described in Japanese Patent Application Laid-Open No. 2002-200002 describes a technique of performing a blurring process on an object boundary portion of a mask pattern for the purpose of suppressing a sense of incongruity. Specifically, in the technique described in Patent Document 1, a first-stage low-pass filter and a second-stage low-pass filter are used to give a slope to the signal at the edge portion of the mask, thereby blurring the mask. . Further, Japanese Patent Application Laid-Open No. 2002-200001 describes adjusting the width of blurring and performing blurring processing for each layer individually (paragraph 0080 of Patent Document 1).

For example, Japanese Patent Application Laid-Open No. 2002-200000 describes a technique for blurring the outline of a designated subject by means of mean filter processing for the purpose of suppressing discomfort. Further, Patent Document 2 describes that the smoothness of the stereoscopic effect is controlled by the size of the average filter (paragraph 0030 of Patent Document 1).

JP 2013-178749 A Japanese Patent Application Laid-Open No. 2003-47027

Here, depending on the two-dimensional image, there may be a portion where the boundary between the foreground and the background is clear and a portion where the boundary is complicated and unclear. If the same blurring process is performed on a location with a clear boundary and a location with an unclear boundary, it may not be possible to effectively suppress the discomfort of the three-dimensional image data.

The aforementioned Patent Document 1 and Patent Document 2 do not mention giving different degrees of blur depending on the location where the boundary between the foreground and the background is clear and the location where the boundary is unclear.

SUMMARY OF THE INVENTION The present invention has been made in view of such circumstances, and an object thereof is to provide an image processing apparatus, an image processing method, and a program for generating three-dimensional image data in which discomfort is effectively suppressed. .

An image processing apparatus according to one aspect of the present invention for achieving the above object is an image processing apparatus for generating three-dimensional image data composed of a foreground and a background from a two-dimensional image. an image acquisition unit for acquiring, a probability calculation unit for estimating a foreground region in a two-dimensional image by image processing and calculating a probability that each small region of the two-dimensional image is a foreground region; A mask generation unit that generates a foreground image mask having a gradation of a mixture ratio of the foreground and the background in a boundary area with the background, and a foreground image acquisition unit that acquires the foreground image using the image mask.

According to this aspect, the probability of being a foreground region is calculated for each small region of the two-dimensional image, and based on the probability, the foreground having a gradation of the mixing ratio of the foreground and the background in the boundary region between the foreground and the background. image mask is generated. As a result, the gradation of the mixing ratio of the boundary region is controlled according to the location where the boundary between the foreground and the background is clear and the location where the boundary is unclear, and the degree of blurring is adjusted.

Preferably, the image processing apparatus includes a background image obtaining section that obtains the background image by interpolating a region corresponding to the foreground image in the two-dimensional image.

According to this aspect, since the background image is obtained by interpolating the area corresponding to the foreground image, it is possible to generate three-dimensional image data in which the sense of discomfort is effectively suppressed.

Preferably, the unmasked portion of the image mask is magnified more than the foreground in the two-dimensional image.

According to this aspect, since the enlarged foreground image can be obtained by enlarging the non-masked portion of the image mask, the missing portion of the background image can be covered, and the sense of incongruity can be suppressed. Image data can be generated.

Preferably, the image processing device includes an extraction region acquisition unit that binarizes the image mask based on the evaluation threshold and acquires the foreground extraction region, and the mask generation unit generates an image composed of the extraction region and the boundary region. Generate a mask.

According to this aspect, the image mask is binarized based on the evaluation threshold, the foreground extraction region is acquired, and the image mask composed of the extraction region and the boundary region is generated. As a result, it is possible to generate three-dimensional image data in which the sense of discomfort is suppressed.

Preferably, the image processing apparatus includes a boundary region acquisition unit that acquires a difference between an enlarged image mask obtained by enlarging the image mask and the extraction region to acquire the boundary region.

According to this aspect, the boundary area is obtained by obtaining the difference between the enlarged image mask and the extraction area, so that the boundary area can be obtained more accurately.

Preferably, the image processing apparatus includes a boundary region obtaining unit that obtains the boundary region by obtaining a difference between an enlarged image mask obtained by enlarging the image mask and a reduced extraction region obtained by reducing the extraction region.

According to this aspect, the boundary area is obtained by obtaining the difference between the enlarged extraction area obtained by enlarging the extraction area and the reduced extraction area obtained by reducing the extraction area. This makes it possible to acquire an enlarged foreground image in which the width of the boundary portion is wider.

Preferably, the border region has a width of no less than 10 pixels and no more than 20 pixels.

Preferably, the probability calculation unit is composed of a trained recognizer for extracting the foreground.

Preferably, the three-dimensional image data is for lenticular printing.

An image processing method according to another aspect of the present invention is an image processing method for generating three-dimensional image data composed of a foreground and a background from a two-dimensional image, the image processing method comprising: obtaining a two-dimensional image; a step of estimating a foreground region in an image by image processing and calculating the probability that each small region of the two-dimensional image is a foreground region; generating a foreground image mask with a blend ratio gradation; and obtaining a foreground image with the image mask.

A program that is another aspect of the present invention is a program that causes a computer to execute an image processing step of generating three-dimensional image data composed of a foreground and a background from a two-dimensional image, the step of acquiring the two-dimensional image. a step of estimating a foreground region in a two-dimensional image by image processing and calculating the probability that each small region of the two-dimensional image is a foreground region; causing a computer to perform image processing steps including generating a foreground image mask having a gradation of mixing ratios of to and background; and obtaining a foreground image from the image mask.

According to the present invention, the probability of being a foreground region is calculated for each small region of a two-dimensional image, and based on the probability, the foreground having a gradation of the mixing ratio of the foreground and the background in the boundary region between the foreground and the background. image mask is generated, the degree of blurring is adjusted by controlling the gradation of the mixing ratio of the boundary region according to the location where the boundary between the foreground and the background is clear and the location where the boundary is unclear.

FIG. 1 is a diagram showing the appearance of a computer. FIG. 2 is a block diagram showing a functional configuration example of the image processing apparatus. FIG. 3 is a flow chart showing the image processing steps. FIG. 4 is a diagram conceptually showing a two-dimensional image and an image mask with probability. FIG. 5 is a diagram showing an enlarged view of the image mask with probability. FIG. 6 is a conceptual diagram showing the gradation of the mixing ratio. FIG. 7 is a diagram explaining extraction of an extraction region. FIG. 8 is a diagram explaining acquisition of the boundary area. FIG. 9 is a diagram illustrating generation of an image mask. FIG. 10 is an enlarged view of the image mask. FIG. 11 is a diagram illustrating acquisition of a foreground image; 12A and 12B are diagrams for explaining the missing portion of the background in the three-dimensional image. 13A and 13B are diagrams for explaining the missing portion of the background in the three-dimensional image. FIG. 14 is a diagram showing a background image. FIG. 15 is a diagram showing images used to generate three-dimensional image data.

Preferred embodiments of an image processing apparatus, an image processing method, and a program according to the present invention will be described below with reference to the accompanying drawings.

FIG. 1 is a diagram showing the appearance of a computer equipped with an image processing apparatus of the present invention.

The computer 3 is equipped with an image processing device 11 (FIG. 2), which is one aspect of the present invention. A two-dimensional image 201 is input to the computer 3 , and a display section composed of a monitor 9 and an input section composed of a keyboard 5 and a mouse 7 are connected. The form of the computer 3 illustrated is merely an example, and a device having functions similar to those of the computer 3 can include the image processing device 11 of the present invention. For example, it is possible to mount the image processing device 11 on a tablet terminal.

The computer 3 displays the two-dimensional image 201 input to the image processing device 11 (FIG. 2) and the generated three-dimensional image data on the monitor 9 . The user inputs commands with keyboard 5 and mouse 7 .

FIG. 2 is a block diagram showing a functional configuration example of the image processing apparatus 11. As shown in FIG. Hardware structures for executing various controls of the image processing apparatus 11 shown in FIG. 2 are various processors as described below. For various processors, the circuit configuration can be changed after manufacturing such as CPU (Central Processing Unit), FPGA (Field Programmable Gate Array), which is a general-purpose processor that executes software (program) and functions as various control units. Programmable Logic Device (PLD), which is a processor, ASIC (Application Specific Integrated Circuit), etc. be

One processing unit may be composed of one of these various processors, or composed of two or more processors of the same type or different types (for example, a plurality of FPGAs, or a combination of a CPU and an FPGA). may Also, a plurality of control units may be configured by one processor. As an example of configuring a plurality of control units with a single processor, first, as represented by a computer such as a client or a server, a single processor is configured by combining one or more CPUs and software. There is a form in which a processor functions as multiple controllers. Secondly, as typified by System On Chip (SoC), etc., there is a form of using a processor that realizes the functions of the entire system including multiple control units with a single IC (Integrated Circuit) chip. be. In this way, various control units are configured using one or more of the above various processors as a hardware structure.

The image processing device 11 includes an image acquisition unit 13, a probability calculation unit 15, a mask generation unit 17, a foreground image acquisition unit 19, a background image acquisition unit 21, a three-dimensional image data generation unit 25, a display control unit 23, and a storage unit 26. Prepare. The storage unit 26 stores programs and information related to various controls of the image processing apparatus 11 . Also, the display control unit 23 controls display by the monitor 9 .

The image acquisition unit 13 acquires the two-dimensional image 201 . A two-dimensional image 201 has, for example, a foreground composed of a main subject and a background other than the foreground.

The probability calculation unit 15 estimates a foreground region in the two-dimensional image by image processing, and calculates the probability that each small region of the two-dimensional image is a foreground region. A known technique is applied to the estimation and probability calculation performed by the probability calculation unit 15 . For example, the probability calculation unit 15 is composed of a learned recognizer for extracting the foreground. The probability calculation unit 15 estimates that each small area is a foreground area by image processing, and calculates a probability based on the estimation for each small area. Also, the small area of the two-dimensional image calculated by the probability calculation unit 15 means an area of a predetermined size, for example, every 1 pixel or every 2×2 pixels.

The mask generator 17 generates an image mask for obtaining the foreground. The image mask has a gradation of the mixing ratio of the foreground and the background in the boundary area between the foreground and the background based on the probability of the area being the foreground calculated by the probability calculation unit 15 . That is, the image mask has a blur based on the probability of being a foreground region in the boundary regions. For the image mask, the probability calculated by the probability calculation unit 15 may be used as it is to configure the masked portion and the non-masked portion. A masked portion and a non-masked portion may be configured with the boundary region. The mask generation unit 17 also has an extraction area acquisition unit 18 and a boundary area acquisition unit 20 .

The extraction area obtaining unit 18 obtains an extraction area. Specifically, the extraction region acquiring unit 18 binarizes the image mask with probability based on the evaluation threshold value and acquires the foreground extraction region. Here, the extraction area is an area that has a certain probability of being a foreground area.

The boundary region obtaining unit 20 obtains a boundary region in the image mask with probability. The boundary area of the image mask with probability has a gradation of the mixture ratio of the foreground and the background.

The foreground image acquisition unit 19 acquires a foreground image using the image mask.

The background image acquisition unit 21 acquires a background image by interpolating a region corresponding to the foreground image in the two-dimensional image.

The 3D image data generation unit 25 generates 3D image data from the foreground image acquired by the foreground image acquisition unit 19 and the background image acquired by the background image acquisition unit 21 . For example, the three-dimensional image data generator 25 generates three-dimensional image data for lenticular printing.

Next, a specific example of generating three-dimensional image data, which is image data for lenticular printing, performed by the image processing apparatus 11 will be described.

<Image processing process>
FIG. 3 is a flow chart showing an image processing process (image processing method) for generating three-dimensional image data from a two-dimensional image by the image processing device 11 . First, the entire image processing process will be described, and then each process will be described in detail.

The image acquisition unit 13 acquires a two-dimensional image A (step S10). After that, the probability calculation unit 15 automatically extracts the foreground region by image processing, and then calculates the probability that it is the foreground. is generated (step S11). Next, the extraction region acquiring unit 18 binarizes the image mask B with probability and acquires the extraction region C (step S12). After that, the boundary region acquiring unit 20 calculates the boundary region based on the extraction region C, and extracts the boundary region D from the image mask B with probability (step S13). Then, the mask generator 17 synthesizes the extraction region C and the boundary region D to generate an image mask E (step S14).

The foreground image obtaining unit 19 applies the image mask E to obtain the foreground image F (step S15). In addition, the background image obtaining unit 21 interpolates the non-masked portion of the image mask E from surrounding pixels to obtain a background image G (step S16). After that, the three-dimensional image data generator 25 generates lenticular printing data using the foreground image F and the background image G (step S17).

Next, each process (step) described above will be described in detail.

<Step S10 and Step S11>
In steps S10 and S11, a two-dimensional image A is input and an image mask B with probability is generated. FIG. 4 is a diagram conceptually showing a two-dimensional image A201 and an image mask with probability B203.

A two-dimensional image A201 has a main subject O that is a person. Three-dimensional image data is generated with the main subject O in the two-dimensional image A201 as the foreground and the background consisting of the portion other than the main subject O. FIG.

The image mask B203 with probability automatically extracts the main subject O by image processing, and has a non-masked portion P corresponding to the foreground portion and a masked portion M corresponding to the background portion. Further, the image mask B203 with probability has a probability of being a foreground region for each small region. The foreground may be acquired by applying the image mask B203 with probability, but by acquiring the foreground by applying the image mask E213 described later, a more accurate foreground image can be acquired.

FIG. 5 is an enlarged view of the image mask with probability B203. 5A shows the image mask B203 with probability, FIG. 5B shows an enlarged view of region H of the image mask B203 with probability, and FIG. , an enlarged view of region J is shown. As indicated by the arrow N in the drawing shown in FIG. 5C, the image mask B203 with probability has a gradation of the mixture ratio in the boundary area between the foreground and the background, and blurs accordingly. have. Note that this gradation is based on the probability that it is the foreground area. In areas where the boundary between the foreground and background is clear, the gradation is steep and narrow, and in areas where the boundary between the foreground and background is unclear. , the gradient becomes looser and wider.

FIG. 6 is a conceptual diagram showing the gradation of the mixing ratio of the boundary portion when the boundary portion between the foreground and background is clear and when the boundary portion between the foreground and background is unclear. FIG. 6A shows the gradation of the probability of being the foreground area when the boundary between the foreground and the background is clear. FIG. 6B shows the gradation of the probability of being the foreground area when the boundary between the foreground and the background is unclear.

In the case shown in FIG. 6A, the area 301 with a 100% probability of being the foreground and the area (background area) 305 with a 0% probability of being the foreground are distinct, so the gradation at the boundary 303 is sharp. narrow. On the other hand, in the case shown in FIG. 6B, since the area 301 with a probability of 100% being the foreground and the area 305 with a probability of 0% being the foreground are unclear, the gradation at the boundary 303 is gentle and wide. . In this way, the gradation of the probability of the foreground region differs between the part where the boundary between the foreground and the background is clear and the part where the boundary is unclear. It is possible to effectively suppress discomfort in the three-dimensional image data.

<Step S12>
At step S12, an extraction region C is acquired. FIG. 7 is a diagram for explaining the extraction of the extraction region C performed by the extraction region acquiring unit 18. As shown in FIG. The extraction region obtaining unit 18 binarizes the probability attached to each small region of the image mask B203 with probability based on the evaluation threshold. For example, when the probability is indicated by 0% to 100%, the extraction region acquiring unit 18 binarizes the region with a probability of 50% or more and the region with a probability of less than 50%. The extraction region C209 has a non-masked portion P with a probability of being a foreground region of 50% or more, and has a masked portion M with a probability of being a foreground region of less than 50%. That is, the extraction region C209 has a binarized non-masked portion P and a masked portion M. FIG.

<Step S13>
In step S13, the boundary area D is obtained. FIG. 8 is a diagram for explaining acquisition of the boundary area D performed by the boundary area acquisition unit 20. As shown in FIG. In the case shown in FIG. 8, the boundary region D211 is generated by subtracting the extraction region C209 from the image mask B203 with probability. Specifically, after enlarging the image mask B203 with probability (enlarged image mask), the boundary area D211 is generated by subtracting the extraction area C209. Alternatively, the boundary area D211 may be generated by subtracting the reduced extraction area C209 (reduced extraction area) after enlarging the image mask B203 with probability (enlarged image mask). For example, after enlarging the image mask B203 with probability by 10 pixels vertically and horizontally, the extraction region C209 is reduced by 10 pixels vertically and horizontally to obtain the difference. The boundary area D211 has a gradation of the mixing ratio of the foreground and background in the boundary area of the image mask B203 with probability. By scaling the image mask B203 with probability and the extraction region C in this way, the width of the boundary region D211 can be controlled. For example, the width of the boundary area D211 is preferably 10 pixels or more and 20 pixels or less.

<Step S14>
At step S14, an image mask E is generated. FIG. 9 is a diagram illustrating generation of the image mask E213 performed by the mask generation unit 17. As shown in FIG. The mask generator 17 obtains an image mask E213 by synthesizing the extraction region C209 and the boundary region D211. For example, after enlarging the image mask B203 with probability by 10 pixels vertically and horizontally, the extraction region C209 is reduced by 10 pixels vertically and horizontally to obtain the boundary region D211. By synthesizing, an image mask E213 is obtained. The image mask E213 has binarized uniform values in the extraction region C209 portion, and has gradation in the boundary region D211 portion. By generating the image mask E213 by synthesizing in this way, the boundary portion between the non-masked portion P and the masked portion M is blurred with a gradation of the mixing ratio of the foreground and the background, and the non-masked portion is blurred. The portion (extraction region) other than the boundary region of P is surely unmasked.

FIG. 10 is an enlarged view of region R of image mask E213 shown in FIG. As shown in FIG. 10, the region R of the image mask E213 has a gradation of the mixing ratio of the non-masked portion P and the masked portion M (indicated by arrow N). By applying the image mask E213 having such a gradation to obtain the foreground and background, it is possible to obtain the foreground and background having appropriate blurring.

<Step S15>
An image mask E is obtained in step S15. FIG. 11 is a diagram illustrating acquisition of the foreground image F215 by the foreground image acquisition unit 19. As shown in FIG. As shown in FIG. 11, by applying an image mask E213 to the two-dimensional image 201, a foreground image F215 is obtained. Since the image mask E213 has a gradation corresponding to the probability that the boundary portion is the foreground area, the foreground image F215 has a blur corresponding to the probability. That is, the degree of blurring of the foreground image F215 is adjusted according to a portion where the boundary between the foreground and the background is clear and a portion where the boundary is unclear.

Here, the foreground image F215 can cover the missing portion of the background by enlarging the main subject O of the original two-dimensional image A201. 12 and 13 are diagrams for explaining the background missing portion in the three-dimensional image. FIG. 12 shows the case where the three-dimensional image is viewed from the front, that is, the case where the foreground image F215 is not moved with respect to the background image. In this case, since the foreground image F215 has not been moved, the missing portion of the background does not appear and the missing portion is not conspicuous. On the other hand, FIG. 13 shows a case where the three-dimensional image is viewed obliquely, that is, a case where the foreground image F215 is moved with respect to the background image. In this case, since the foreground image F215 has been moved, the missing portion U of the background is generated and the missing portion U stands out. In such a case, the missing portion U can be covered by enlarging the foreground image F215.

<Step S16>
At step S16, a background image G is obtained. FIG. 14 is a diagram showing the background image G217 acquired by the background image acquisition unit 21. As shown in FIG. In the diagram shown in FIG. 14, the background image acquisition unit 21 generates a background image G217 by complementing the foreground image F215 shown in FIG. Note that the complementing performed by the background image acquisition unit 21 is performed by a known technique.

<Step S17>
In step S17, data for lenticular printing is generated. FIG. 15 is a diagram showing images used to generate three-dimensional image data. The image 223 is a front view, and the positional relationship between the foreground and the background is the same as that of the original two-dimensional image A. FIG. Image 221 moves foreground image F215 in the direction of arrow V with respect to background image G217, and image 225 moves foreground image F215 in the direction of arrow W with respect to background image G217. For example, the three-dimensional image data generator 25 uses the images 221, 223, and 225 to generate three-dimensional image data for lenticular printing.

Each configuration and function described above can be appropriately realized by arbitrary hardware, software, or a combination of both. For example, a program that causes a computer to execute the above-described processing steps (procedures), a computer-readable recording medium (non-temporary recording medium) recording such a program, or a computer capable of installing such a program However, it is possible to apply the present invention.

Although examples of the present invention have been described above, it goes without saying that the present invention is not limited to the above-described embodiments, and that various modifications are possible without departing from the spirit of the present invention.

3 : Computer 5 : Keyboard 7 : Mouse 9 : Monitor 10 : Vertical and horizontal 11 : Image processing device 13 : Image acquisition unit 15 : Probability calculation unit 17 : Mask generation unit 18 : Extraction region acquisition unit 19 : Foreground image acquisition unit 20 : Boundary Area acquisition unit 21 : Background image acquisition unit 23 : Display control unit 25 : Three-dimensional image data generation unit 26 : Storage unit 201 : Two-dimensional image

Claims (12)

An image processing device that generates three-dimensional image data composed of a foreground and a background from a two-dimensional image,
an image acquisition unit that acquires the two-dimensional image;
a probability calculation unit that estimates the foreground region in the two-dimensional image by image processing and calculates a probability that each small region of the two-dimensional image is the foreground region;
An image mask composed of a non-masked portion corresponding to the foreground portion and a masked portion corresponding to the background portion for the two-dimensional image, wherein the foreground and the a mask generation unit for generating an image mask for obtaining the foreground, which has a gradation of a mixing ratio of the foreground and the background in the boundary area with the background , based on the probability;
a foreground image obtaining unit that obtains a foreground image using the image mask;
with
In the mixing ratio, the higher the probability, the higher the ratio of the foreground to the background.
Image processing device. 2. The image processing apparatus according to claim 1, further comprising a background image acquisition unit that acquires a background image by interpolating a region corresponding to the foreground image in the two-dimensional image. 3. The image processing apparatus according to claim 1, wherein the non-masked portion of the image mask is magnified more than the foreground in the two-dimensional image. an extraction region acquisition unit that binarizes the image mask based on an evaluation threshold and acquires the foreground extraction region;
4. The image processing apparatus according to any one of claims 1 to 3, wherein the mask generation unit generates the image mask composed of the extraction region and the boundary region. 5. The image processing apparatus according to claim 4, further comprising a boundary area acquiring unit that acquires a difference between an enlarged image mask obtained by enlarging the image mask and the extracted area to acquire the boundary area. 5. The image processing apparatus according to claim 4, further comprising a boundary area obtaining unit that obtains the boundary area by obtaining a difference between an enlarged image mask obtained by enlarging the image mask and a reduced extraction area obtained by reducing the extraction area. 7. The image processing device according to any one of claims 1 to 6, wherein the border area has a width of 10 pixels or more and 20 pixels or less. The image processing apparatus according to any one of claims 1 to 7, wherein the probability calculation unit is configured by a learned recognizer for extracting the foreground. 9. The image processing apparatus according to claim 1, wherein said three-dimensional image data is for lenticular printing. An image processing method for generating three-dimensional image data composed of a foreground and a background from a two-dimensional image,
obtaining the two-dimensional image;
estimating the foreground region in the two-dimensional image by image processing, and calculating the probability that each small region of the two-dimensional image is the foreground region;
An image mask composed of a non-masked portion corresponding to the foreground portion and a masked portion corresponding to the background portion for the two-dimensional image, wherein the foreground and the generating an image mask for obtaining the foreground, having a gradation of the mixing ratio of the foreground and the background in the boundary region with the background , based on the probability;
obtaining a foreground image with the image mask;
including
In the mixing ratio, the higher the probability, the higher the ratio of the foreground to the background.
Image processing method. A program for causing a computer to execute an image processing process for generating three-dimensional image data composed of a foreground and a background from a two-dimensional image,
obtaining the two-dimensional image;
a step of estimating the foreground region in the two-dimensional image by image processing, and calculating the probability that each small region of the two-dimensional image is the foreground region;
An image mask composed of a non-masked portion corresponding to the foreground portion and a masked portion corresponding to the background portion for the two-dimensional image, wherein the foreground and the generating the image mask for obtaining the foreground, which has a gradation of the mixing ratio of the foreground and the background in the boundary region with the background , based on the probability;
obtaining a foreground image with the image mask;
including
A program for causing a computer to execute an image processing step, wherein the mixing ratio is such that the higher the probability, the higher the ratio of the foreground to the background . A non-transitory computer-readable recording medium, wherein when instructions stored on the recording medium are read by a computer,
An image processing step for generating three-dimensional image data composed of a foreground and a background from a two-dimensional image,
obtaining the two-dimensional image;
a step of estimating the foreground region in the two-dimensional image by image processing, and calculating the probability that each small region of the two-dimensional image is the foreground region;
An image mask composed of a non-masked portion corresponding to the foreground portion and a masked portion corresponding to the background portion for the two-dimensional image, wherein the foreground and the generating an image mask for obtaining the foreground, which has a gradation of the mixing ratio of the foreground and the background in the boundary region with the background , based on the probability;
obtaining a foreground image with the image mask;
including
A recording medium for causing a computer to execute an image processing step, wherein the mixing ratio is such that the higher the probability, the higher the ratio of the foreground to the background .

Images