JPH0591545A - Stereoscopic image recording/reproducing system - Google Patents

Abstract

PURPOSE:To make it possible to reproduce a three-dimensional image without depending upon a binocular parallax method by dividing an original image in each hierarchy in the depth direction, rerecording the divided images and successively reproducing the rerecorded contents as real images having respectively different space forming positions. CONSTITUTION:The distance of an object is extracted from the outlines or the like of geometric elements of the object included in a reference frame to find out the depth of the object and the object is divided in the picture depth direction based upon the found depth to find out the number of divided planes. The divided object is projected to the nearest divided plane to obtain a projection image projected to the divided plane and the projection image is recorded in an image recording medium 11 together with depth position information expressing the position of the divided plane in the picture depth direction by an identical distance image extracting/recording means 9. Thus the object in the image is stereoscopically expressed from the plane image. Consequently an already recorded planar image can be converted into a stereoscopic image without using a specific stereoscopic photographing device, the stereoscophic image can be rerecorded and reproduced and the stereoscopic image having optional screen size can be provided.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image source of a stereoscopic image device for projecting a stereoscopic image of a two-dimensional image such as an image captured by an image capturing device such as a television camera or an image reproduced from an image recording medium. Image conversion to convert to.

[0002]

2. Description of the Related Art Conventionally, various attempts have been made to provide images with a sense of depth, so-called stereoscopic images. Although various methods have been proposed for forming a stereoscopic image, a method that uses so-called binocular parallax is often used because of its ease of application to moving images and production. In this method, one object is photographed from two directions corresponding to both eyes of a person, and each image is presented to the corresponding eye of the observer by an appropriate method, thereby giving the observer a stereoscopic effect. Let

Since this binocular parallax system requires images from two directions, two photographing devices are prepared for photographing.
Two photographed images are simultaneously photographed and recorded from two directions corresponding to both eyes. Therefore, the recorded image depends on the installation condition of the image capturing device at the time of image capturing, and the observation condition of the person who observes the reproduced image also needs to be adjusted to this image capturing condition. For example, if an image taken on the assumption that it is provided as a small screen is projected on a large screen, and if the distance between the observer and the screen does not differ much from the shooting conditions, two images for the left and right eyes are displayed.
There is a problem that two images are not separated into three-dimensional images because they are relatively separated from each other and are observed as double images. On the contrary, when observing the image taken on the small screen assuming that it is provided as a large screen,
When the distance between the observer and the screen is not so different from the shooting condition, parallax is hardly recognized and a stereoscopic effect is not generated.

Such a problem is common not only to the binocular parallax system but also to the stereoscopic image observing system requiring a plurality of parallaxes.

[0005]

As described above, in the conventional image reproducing apparatus for stereoscopic vision, a plurality of planar images photographed by the photographing apparatus dedicated to stereoscopic images are individually recorded,
It is necessary to play each image under the same conditions that were set at the time of shooting during playback, and if the size of the image playback screen changes, the image will not be reproduced with appropriate parallax, resulting in a three-dimensional effect. It is difficult to obtain enough. In addition, a photographing device dedicated to stereoscopic images is not suitable for general use because of difficulty in setting the device, size of the device, and the like.

Therefore, according to the present invention, a stereoscopic image conversion method and a stereoscopic image capable of forming a stereoscopic image from image information recorded as a planar image and reproducing a three-dimensional image without depending on the binocular parallax system. An object is to provide a conversion device.

[0007]

In order to achieve the above object, a stereoscopic image recording apparatus which constitutes a stereoscopic image recording / reproducing system of the present invention forms a stereoscopic image from one basic frame of an image signal having basic frame intervals. In the stereoscopic image recording apparatus for forming an additional frame of, and inserting this between the basic frames to form a stereoscopic image signal and recording the stereoscopic image signal, the outline of the geometrical elements of the subject included in the basic frame A means for obtaining the depth of the subject by extracting the perspective of the subject, and obtaining the number of division planes for dividing the subject in the depth direction of the screen based on the depth, and a division plane for projecting the divided subject on the nearest division plane. The same distance image is obtained by obtaining the projection image projected on the image recording medium and recording it on the image recording medium together with depth position information indicating the position of the division plane in the screen depth direction. Characterized in that it comprises an extraction recording means.

Further, in the stereoscopic image reproducing apparatus which constitutes the stereoscopic image recording / reproducing system of the present invention, the image signal for carrying the two-dimensional image and the depth position signal representing the position where the two-dimensional image should be projected are recorded. An image recording medium playing means for playing the image recording medium to demodulate the image signal and the depth position signal, and an image forming position by driving the image projection lens in the optical axis direction in response to the supply of the depth position signal. And a projection lens control means for controlling the image signal, and an image projection means for converting the image signal into a two-dimensional image and projecting the two-dimensional image via the image projection lens.

[0009]

The three-dimensional image recording device extracts geometric elements from a two-dimensional image and estimates the depth of the subject. For such estimation, for example, a subject is discriminated from geometrical elements extracted from the screen by pattern recognition, the three-dimensional shape data of the subject is read from a database of various subject shapes, and the depth of the subject is estimated from the shape data. Is possible.
When there are a plurality of subjects, it is possible to estimate the distance relation in the depth direction of the two-dimensional subjects in consideration of the positional relation between the subjects. Based on the estimated depth of the screen, the number of division planes that divide the subject into planes perpendicular to the depth direction is obtained, and the division planes are arranged at a predetermined distance in the depth direction of the screen to divide the subject. The divided subject is projected on the dividing surface, and the image projected on the dividing surface is inserted as an additional frame between the basic frames of the original image together with the position in the depth direction of the dividing surface and recorded on the image medium. The stereoscopic image reproducing device projects a series of image frames at high speed while setting the projection position of the reproduced image for each frame.

As a result, it is possible to observe the recorded planar image information which is not photographed for stereoscopic vision as a stereoscopic image. In addition, even in the case of an image captured and recorded for stereoscopic viewing, even if there is a difference in observation conditions such as a difference in screen size during reproduction, a distance between the screen and an observer, etc. It can be corrected and can be recorded or reproduced as an appropriate stereoscopic image.

[0011]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The principle of stereoscopic vision according to the stereoscopic vision apparatus of the present invention will be described with reference to FIG. FIG. 5A shows a state in which two cubes A and B, which are subjects, are recorded as a plane image. A procedure for forming the cubes A and B as a stereoscopic image will be described.

First, a geometric figure forming an image is extracted from a plane image, and what kind of figure is extracted is discriminated by using a technique such as pattern recognition. For this reason, the contents of the planar image source are checked in advance, and the type of subject appearing on the screen and its three-dimensional model are accumulated to construct a database. A subject is discriminated from the extracted figure, and a three-dimensional model is selected. This three-dimensional model is placed in a virtual three-dimensional space formed in the computer shown in FIG. The arrangement can be set from the contour of the geometric figure by using the perspective method or the like. If the unit model registered in the database is a solid element, cubic solid elements A and B are arranged in the virtual space. If the front direction of the screen is made to correspond to the X-axis direction of the virtual space, the solid elements A and B can be calculated by the data representing the shape of the solid model.
Mutual depth is revealed. For example, among the P points (x, y, z) forming the surface of each model, the depth distance of the model group is obtained from the difference between the maximum x _max and the minimum x _min . The solid element is divided by the YZ plane according to this depth. If the depth is large, the number of divisions is increased. In FIG. 5B, l
It is divided by three planes of ₁ to l _3. This number of divisions is related to the resolution of the stereoscopic image reproduced later. Then, the part where the solid element projects from the divided plane is obtained. S _{1 in} FIG. 5 (C) indicates a portion of the solid element protruding forward from the plane l ₁ . S ₂ represents a solid element lying between the planes l ₁ and l ₂ . S ₃ is the plane l ₂ and l
_The solid elements that are present between ₃ are shown. S ₄ is the plane l ₃
The solid-state elements existing between the background and a plane l _{∞ (} not shown) serving as a background are shown. Obtaining additional frames F ₁₁ to F ₁₄ thus solid elements S ₁ to S ₄ which is divided by a plane l ₁ to l ₃ projecting respectively from the front direction to the plane l ₁ to l _∞.
That is, the additional frames F _{11 to} F ₁₄ are obtained by displaying the pixels of the surface viewed from the front other than the cut surface of the divided solid elements S _{1 to} S _{4 with} the YZ components, respectively. In addition to the image information, the additional frames F _{11 to} F ₁₄ also include frame position information x ₁ ′ to x _{1 ′} representing the frame position to be projected.
₄ 'is added. Background frame F ₁₀ to remove solids components A and B from the basic frame F ₁₀ to the frame to form the three-dimensional image of this like, the frame F ₁₀ by adding positional information x _∞
One frame of stereoscopic screen information including F ₁₄ and F _{10 ∞} is formed. When such image processing is repeated for each frame f _{7 to} f ₁₂ of the original image of FIG. 6B, FIG.
A series of stereoscopic image signals shown in is obtained. This is configured such that additional frames indicated by dotted lines for forming a stereoscopic image are inserted between frames of the original image signal indicated by solid lines in a number according to the required stereoscopic resolution.

The video signal representing each frame includes, for example, a frame number as an address of the frame, a time code indicating a lapse of projection time from the initial frame, and position information indicating a position where the frame should be projected corresponding to the depth of the subject. Inserted between frames.

The stereoscopic image signal is recorded on an image recording medium such as a video disc or a video tape.

In a stereoscopic image reproducing apparatus not shown,
The position information recorded for each frame is demodulated and the position where each frame should be projected is set by the movable projection lens of the projector or the mirror for changing the optical path length, and the video signal is demodulated to project the image of each frame. To form a series of images that are projected and moved toward the front of the screen. When an observer views a series of images in which the projection position changes at a high speed in the front direction of the screen, the series of planar images is recognized as a stereoscopic image due to the afterimage effect of vision.

An embodiment of the present invention will be described with reference to FIGS. 1 and 2. FIG. 1 shows the configuration of a stereoscopic image conversion system of the present invention, and FIG. 2 shows the configuration of a stereoscopic image reproduction system.

Video disc, video tape, CD-R
An original image is recorded on the image recording medium 1 such as an OM and a hard disk. The image recording medium 1 is played by an image reproducing device 2 such as a video disc player or a digital VTR, and the number of frames corresponding to the capacity of the image memory in the subsequent stage, for example, one frame is reproduced. Each reproduced frame is converted into a digital value that carries color information of the brightness and hue of each pixel and stored in the image memory 4. For example, if one pixel is displayed with 8 bits, 256 colors can be represented. Further, it is possible to convert a video signal obtained from the photographing device 3 such as a television camera into a digital value and store a predetermined frame in the image memory 4.

The image memory 4 gives an image for one frame to the depth estimation device 5. The depth estimation device 5 is a computer suitably configured for image data processing, and forms the above-described three-dimensional virtual space in a memory (not shown). Then, the image recognition unit 5 for identifying what is the geometrical characteristic value of the subject formed in the image memory 4
a, an overall position estimating unit 5b that estimates the front-rear relationship between the identified subjects, and an individual position estimating unit 5c that estimates the overall positional relationship between the subjects. Further, it is also possible to make the same distance image extraction recording device 9 described later have an image extraction function. These estimations are executed by sequentially referring to the data stored in the database device 6 in which the general size of the subject and the basic data relating to the geometrical relation are registered.

Even if a database in which the shape of a subject is registered in advance, the data can be appropriately corrected by a manual operation by an operator when the degree of conformity indicating the accuracy of subject discrimination is low and estimation is difficult. The data correction device 8 is used. The interface with the operator is performed by the data input terminal device 10 provided with a keyboard, an information display device such as an image, and the like. The corrected data is registered as new data in the database via the database correction device 8. Further, the new data obtained from the above estimation result is registered as new data in the database by the data updating device 7.

The estimation result by the depth position estimating device 5 is sent to the same distance image extracting and recording device 9, and the original image is corrected and added according to the type and characteristics of the image reproducing device 13 described later. The same-distance image extraction recording device 9 is a computer equipped with an image recording device and suitable for image processing. The part belonging to is extracted to form one screen from the original screen, and this is recorded in the image recording medium 11. A continuous video signal is formed by repeatedly forming a plurality of additional frames from one frame of the original image and sequentially recording the background frame and the additional frame group on the image recording medium 11.

When the image reproducing device 13 also serves as a binocular parallax stereoscopic device, the parallax of the left and right binocular images and the mutual distance are determined by the screen size for presenting the stereoscopic image and the distance to the observer. Correct the value accordingly and record it.

Next, the operation of the depth position estimating device 5 will be described with reference to the flow chart shown in FIG. As described above, the main part of the depth position estimating device 5 is composed of a computer (hereinafter, abbreviated as CPU), and can also control related devices. CPU
When the device starts image processing to obtain a stereoscopic image from a two-dimensional image, the image recording medium reproducing device 2 or the photographing device 3 playing the original image recording medium 1 stores image information in the image memory 4 ( In step S103), the following process is executed.

First, a step of recognizing the shape of an object displayed on the screen (steps S104 to S113), a step of detecting a contour of the object and extracting a solid figure as a figure element (steps S104 to S106), A step of examining the combined state of the plurality of extracted graphic elements and estimating them as independent solid elements (steps S107 to S10).
8), a step in which a person checks a portion having a low degree of conformity in the automatic discrimination (steps S109 to S110), a step of naming the checked solid element (step S111), a step of registering these in a database (Steps S112 to S113) are executed. Each step includes a process of determining the completion status of the above-described operation and a supplementary operation of performing a process for determining these when the determination is negative.

In the contour extraction process (S104), for example, a region gathered by the lightness distribution or color distribution of the image is extracted, edges are emphasized from these changes (step S105), and then the basic figure shape or A publicly known method for determining the contour by the interpolation curve (step S106) is adopted.

For the combined state (S107), the inclusion relation of each graphic element determined in the previous stage is estimated from the geometry and the actual appearance of the object (step S108), and the solids including these graphic elements are estimated. Elements are determined. The portion of which the estimation fitness is low and cannot be determined (step S109),
The operator can perform the correction work from the data input terminal 10 via the database correction device 8 (step S110). When the operator names or labels graphic elements (S111), the data input terminal 10 or the like is used (S110). When all have been determined, the data is registered in the database device 6 (S112), but the database updating work (step S113) at this time is performed by the data updating device 7. These operations by the operator correspond to the complement of the function of the image recognition unit 5a shown in FIG. 1 or a kind of learning of the image recognition unit 5a.

Subsequently, the CPU determines the front and rear positions of the solid element on the screen (steps S121 to S13).
0) is executed. This is a part of the function of the overall position estimation unit 5b. These steps are also composed of a process of judging and a process when the judgment is negative. That is, the process of discriminating the context from the graphic element shape itself (step S12
1), a determination process in relation to ambient light such as shadow, lightness, and hue (step S123), and an estimation process based on optical conditions at the time of shooting such as focus / blur degree at the time of shooting (step S12)
5), check process by operator (step S12)
7), database registration process (step S129)
Is equipped with. In each of these processes, if the conformity of the judgment result is low, a supplementary operation is performed.

First, in the front-back determination based on the shape (S121), geometric knowledge such as overlapping and continuity of geometric figures is extracted from the database device 6, and an estimation operation is performed in comparison with the original image (step). S122). Environmental conditions (S
In 123), a knowledge database (step S124) regarding the reflection of light, the scattered state, and the degree of bending of the shadow is used. A database of the degree of focusing of the lens at the time of shooting and the relative relationship in the depth direction of the screen (step S1.
26) is also useful. If the operator needs to perform a check or a database update operation (step S127), the same steps as those described above are performed (steps S128 and S130).

Next, a step of quantifying the context of each solid element obtained in the above processing (steps S131 to S14)
1), a process of extracting a figure having standard dimensions (step S13)
2), a perspective vanishing point extraction step (step S13)
4) The process of deriving the degree of separation by light, shadow, or color (step S136) and the checking / registering process (steps S138 and S140) are executed. Step S131-
S141 is a part of other functions that the overall position estimation device 5b has.

When the quantification work has already been completed (S1
31) shifts to the next stage through the check / registration work and the additional work (steps S139 and S141) similar to those described above. If not (S13
1) If the size of the solid element is not extracted (S13)
2) Comparing the registered solid elements with the knowledge database, the standard size solid elements (mountain, sign, building,
People, industrial products, etc.) are searched (step S133). Next, when the vanishing point is not extracted (S134), the vanishing point is obtained from the deformation state of the already registered object in one screen, and the geometrical characteristics and the shape of the solid element are compared, The distance relationship between each solid element is estimated based on the perspective method (step S135). If the quantification based on the shadow or the like has not been performed (S136), it is possible to perform the quantification from the extent of shadow expansion, the extent of color fading, or the attenuation of light amount (step S137). When the degree of conformity of the estimation result in the above steps is low (S138), the operator may correct data or input new data via the data input terminal device 10 or the database correction device 8 (S1).
39).

Next, a step of quantifying the context of the graphic elements included in each solid element (steps S151 to S15).
Execute 9). For example, those relating to the unevenness of a solid element such as the height of the nose of a human being and the degree of swinging up of limbs are common. These series of steps are mainly judgment operations, and the contents thereof are substantially the same as the above-described quantification process of the context of the solid element. The different process is the vanishing point extraction process (step S
In 134), it is not necessary to do it here because it has already been requested. This step (S151 to S159) corresponds to the function of the individual position estimation unit 5c.

Although the steps described so far are for one screen, it is difficult to make a judgment with only one screen and a database for the previous and next screens is required (step S16).
In 1), call another screen to the image memory 4 and
Alternatively, when other screens are also stored in the image memory 4, the above-described various processes are performed on the other screens (S
162), and improves the adaptability of the estimation work by comparing the result with the previous result (S163). The same checking and registration work as above (step S164)
~ S167) is attached.

After the above steps are completed, the number of depth screens (additional frames) and the position of the depth screen are determined based on the depth resolution of the image reproducing apparatus 13 actually used and the depth of the screen (step S171). FIG. 5C shows solid elements / graphic elements corresponding to the number and position of the depth screens.
As described above, it is re-projected on the depth screen that divides the solid element / graphic element and recorded in a reproducible state on an appropriate image recording medium by the same distance image extraction recording device 9 (step S17
2). For each frame of the original image, step S103-
By repeating S172, the stereoscopic image source converted from the stereoscopic image source to the stereoscopic image projection is obtained on the image recording medium 11. In this stereoscopic image source, performance information such as depth information, frame number, and time code is inserted for each frame, for example, in a blanking interval area between frames.

When generating a stereoscopic image, the image recording medium 11 is played by the reproducing device 12, the depth information is demodulated together with the recorded image information, and this is used as a video signal and a projection position control signal. Supply to 13. The image reproducing device 13 forms a screen based on the image information of the video signal, and sets the projection position of the screen for each frame in the depth front direction of the image projection space based on the depth information of the projection position control signal. By rapidly changing the formation position of each screen in the front direction, a group of images that remain as a visual afterimage of human beings are combined, and the observer looks as if a stereoscopic image exists.

Note that the depth resolution (screen density in the X-axis direction) of the depth screen in the front direction is higher as it is closer to the observer in accordance with the visual characteristics, and those beyond a certain depth distance are only the background flat image. can do.

FIG. 7 shows an example of the configuration of a stereoscopic device using the device for observing a stereoscopic image while continuously reproducing images at the same distance among the image reproducing devices 13 described above.

The video signal demodulated by the image recording medium playing device 12 is used as an image correction unit 13 of the image reproducing device 13.
sent to a. The image correction unit 13a finely adjusts the distortion, pedestal level, etc. of the projected image according to the characteristics of the image reproduction system that actually uses the video signal. Output. An image signal portion of this output signal is sent to an image projection device 13b such as a projector CRT and converted into an optical image. This optical image is a movable lens device 1 that is moved by a motor.
By 3c, the established position of the real image 14 can be set to arbitrary points 14a and 14b in the depth direction. Movable lens controller 13
d is the movable lens device 13c based on the projection position control signal.
Drive the voice coil motor. As a result, the projection lens moves forward and backward along the optical axis at high speed, and the position of the real image 14 formed in the space moves forward and backward in the depth direction. When the video signal and the projection position control signal are demodulated and supplied as a composite signal, the signal separation unit 13i separates the video signal carrying the image information and the projection position control signal carrying the lens control information.

The vertical position of the real image formed in the space may not be adjusted by the image correction unit 13a or the movable lens unit 13c, but in this case, two reflecting mirrors 1 are used.
The established position of the image can be changed using 3e and 13g. The reflecting mirrors 13e and 13g are respectively the reflecting mirror attitude control device 13f.
And 13h, the operation with two degrees of freedom is performed for each reflecting mirror.

The projected image from the image reproducing device 13 is formed along the image optical axis 15, and passes through the concave mirror 17 to the observer 16
Will be presented to. The concave mirror may have a form in which only the upper and lower parts 17a and 17b of the concave mirror are left as shown in the figure, a form in which the entire surface of the concave mirror is left, or a form in which the concave surface is formed by a large number of small plane mirrors. If the positions of the above-mentioned projected images are 18a and 18b, the reflected images formed by the concave mirror will be 14a and 14b.

In the present example, the image optical axis 15 and the optical axis 19 of the concave mirror are set to be parallel, but they may be set to have an angle to some extent. At this time, the respective units in the image reproducing device 13 are reset to proper states.

FIG. 8 shows an example in which the image temporarily recorded in the above-mentioned image memory is an image for the left and right eyes for binocular stereoscopic vision. Actually, only one image for one eye is recorded on one screen of the image memory, but for the sake of convenience, when an image for both eyes is drawn at the same time, it is shown as a double image separated by l ₁ to the left and right like 21a. Be done. In the figure, for simplicity of explanation, the image is described as a circle. This distance l ₁ needs to correspond to the human pupil distance, but when the size of the presentation screen increases,
The size of the image also becomes large like 21b, and as a result, the separation distance l ₂ between the left and right eye images also increases, and it does not match the pupil distance. On the contrary, when the size of the presentation screen is reduced to 21c, the separation distance l ₃ between the left and right eye images is reduced, and in this case also, it does not match the pupil distance. In any case, there is a high possibility that the conditions are unsuitable for observing a stereoscopic image. Therefore, the image in which the separation distance between the left and right eye images is corrected to the appropriate distance l ₁ according to the size of the presentation screen is recorded in the same distance image extraction recording device 9. In this case, the images to be recorded are not the images of the same distance in the strict sense, but the depth position estimating device 5 is used as in the above example. That is, by simply adjusting the separation distance between the left and right eye images,
Since distortion may occur in the reproduced stereoscopic image,
After the image recognition unit 5a and the position estimation unit 5b estimate the three-dimensional shape of the observation target, the individual position estimation unit 5c generates left and right eye images when observed with an appropriate parallax. In this generation, a parallax image corresponding to the other part of the original image is searched and used, or a new image is generated by using the previous estimation result.

The contents described above are an example of the apparatus according to the present invention, and various modifications and applications such as, for example, changing the order of the processing steps, introducing other known methods for achieving the purpose of the processing steps, etc. Is possible.

[0042]

As described above, according to the present invention, since the subject in the image is three-dimensionalized from the two-dimensional image, a special stereoscopic photographing device is not required and the already recorded two-dimensional image is stereoscopically viewed. It can be converted into an image for recording, re-recorded, and reproduced. In addition, it is possible to present a stereoscopic image with a free screen size without allowing an observer to wear special glasses for stereoscopic viewing. Furthermore, it is also possible to apply to an image recorded for stereoscopic viewing.

[Brief description of drawings]

FIG. 1 is a block diagram showing a conversion system for stereoscopic image conversion according to the present invention.

FIG. 2 is a block diagram showing a stereoscopic image reproducing system.

FIG. 3 is a flowchart showing an image processing procedure.

FIG. 4 is a flowchart showing an image processing procedure.

FIG. 5 is a diagram illustrating a state in which a relationship in the depth direction of a planar image is extracted and re-recorded.

FIG. 6 is a diagram illustrating a stereoscopic image signal.

FIG. 7 is a diagram showing a reproduction system for generating a stereoscopic image.

FIG. 8 is a diagram for explaining correction of an image recorded for binocular parallax.

[Explanation of symbols]

 1, 11 Image recording medium 4 Image memory 5 Depth position estimation device 9 Same-distance image extraction recording device 12 Image recording medium playing device 13 Image reproducing device

Claims (3)

[Claims] 1. A geometric element is extracted from plane image information including at least one subject arranged in a three-dimensional space, depth information representing the perspective of the geometric element is extracted from the plane image information, and The object corresponding to the geometrical element extracted by referring to the database in which the relationship between the geometrical element and the object is recorded is estimated, and the object of each object estimated in advance from the database in which the three-dimensional shape of the object is recorded is estimated. The three-dimensional shape is read and placed in a virtual three-dimensional space, the positional relationship in the depth direction of each subject is estimated based on the depth information, and the number of additional frames is calculated based on the positional relationship. The three-dimensional shape placed in is divided by the number of additional frames by the plane perpendicular to the depth direction to obtain the divided three-dimensional shape, and each divided three-dimensional shape is the closest three-dimensional shape. A three-dimensional image characterized by recording on a recording medium together a plane image group obtained by projecting on each of the divided planes and a depth information group indicating the position of each of the divided three-dimensional shapes in the depth direction. Recording method. 2. A stereoscopic image forming an additional frame for forming a stereoscopic image from one basic frame of an image signal having basic frame intervals, and inserting the additional frame between the basic frames to form and record a stereoscopic image signal. An image recording device, wherein the depth of a subject is obtained by extracting the perspective of geometrical elements of the subject included in the basic frame, and the number of division planes for dividing the subject in the screen depth direction is determined based on this depth. And a means for obtaining the projection image of the divided object on the nearest division plane to obtain a projection image projected on the division plane, and image recording the same with depth position information indicating the position of the division plane in the depth direction of the screen. A stereoscopic image recording device comprising: same-distance image extraction recording means for recording on a medium. 3. An image recording medium on which an image signal carrying a two-dimensional image and a depth position signal representing a position where the two-dimensional image should be projected are recorded to play the image signal and the depth position signal. Image recording medium playing means for demodulating, projection lens control means for controlling the image forming position by driving the image projection lens in the optical axis direction in response to the supply of the depth position signal, and the image signal into a two-dimensional image. An image projecting means for converting and projecting the image through the image projection lens.