JP2007272318A - Processing method of personal photographic image - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To obtain a portrait photographic image with good appearance by processing a portrait photographic image taken by a camera from the front of a person, to slightly change the direction. <P>SOLUTION: The processing method of personal photographic image for changing the direction of the person in a portrait picture by use of a portrait photographic digital image taken from the front of the person comprises steps of cutting a portrait image area 2 from the portrait photographic digital image 1 to separate the portrait image area 2 from a background image 3 of the person; pasting the separated portrait image area 2 to a rotary model 4; rotating the rotary model 4 by a designated rotating angle; pasting the portrait image area 2 in the rotary model 4 changed in the direction by the rotation to the separated background image 3 to reconstitute the portrait photographic digital image 1, whereby the direction of the person in the portrait picture is changed. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

The present invention provides a digital image of a person photograph taken from the front of a person by a mobile phone with a camera or a digital camera, or a means such as a scanner for taking a photograph of a person taken from the front of a person by an ordinary camera (film camera). The present invention relates to a processing method for processing a digital image of a human photograph digitized by the method to obtain a human photograph image having a good appearance.

2. Description of the Related Art Conventionally, there has been known a processing method for obtaining a person photo image having a good appearance by changing the contrast of light and shade of a subject image or changing the color tone. Also, as disclosed in Patent Document 1, there is also known a photo processing method in which a subject image is deformed in a plane to make it a solid physique or a chubby physique. Furthermore, a technique for changing facial expressions and movements by image processing is also known.

On the other hand, when taking a picture of a person, if the person is taken from an angle rather than from the front, the appearance will be better. In particular, it is said that a woman's face looks most beautiful when taken from an oblique left side. However, when humans are pointed at the camera, they tend to unconsciously hold the camera from the front. For this reason, in many cases, a photographic image of a person is a front image, and the visual impression is not good.

For this reason, image processing is effective in which a person in a photographic image taken in front of the camera is directed diagonally. For this purpose, a direction conversion function is required for the person in the course of image processing. It is clear that this orientation conversion function cannot be realized simply by changing the contrast of the shade of the subject image or changing the color tone, and by transforming the subject image in the plane, It cannot be realized by a photo processing method that makes a chubby physique, nor can it be realized by a technique that changes the facial expression or movement of the main part of the face.
JP 2000-267156 A

Accordingly, an object of the present invention is to obtain a person photo image having a good appearance by processing a photograph image of a person photographed by a camera from the front of the person and directing the person obliquely.

Under the above-described problems, the present invention separates a person image area from a person background image by cutting out the person image area from the person photograph digital image in a digital image of a person photograph taken from the front of the person. The process of pasting the separated human image area onto the rotating body model, the process of rotating the rotating body model by the specified rotation angle, and the image area of the person on the rotating body model whose orientation has been changed by the rotation. By pasting it on the background image separated in the process and reconstructing the digital image of the person photo, the direction of the person in the person photo is changed to obtain a person photo image having a good appearance.

In the present invention, the rotator model has a size corresponding to the size of the upper body of the person, and a solid formed by rotating the center line of the upper body area of the cut out person is used as the rotator model. (Claim 2). Further, according to the present invention, in the background image after the person image area is cut out, the area where the person image area exists is filled with a partial image of the background image.

In the present invention, the person image area is cut out from the digital image of the person photograph taken from the front of the person, the image area of the person is pasted on the rotating body model, the rotating body model is rotated by a specified rotation angle, A digital image of a person photograph is reconstructed by pasting the image area of the person in the rotating body model whose direction is changed by the rotation to the separated background image. For this reason, since a person photograph (face photograph) from a good looking direction can be obtained by image processing operations based on a person photograph (face photograph) photographed from the front of the person, it is not necessary to shoot again. It is possible to create a person photo (face photo) with good appearance (claim 1).

Since the rotating body model uses a rotating body that has a size corresponding to the size of the upper body of the person and is rotated around the center line of the upper body area of the person, the matching of the upper body of the person and the rotating body model is combined. The upper body of a person having a specified rotation angle obtained by rotation of the rotating body model, in particular, the face is not unnatural and is advantageous (claim 2).

In addition, in the background image after cutting out the person image area, the area where the person image area was filled is filled with a partial image of the background image. (Claim 3).

FIG. 1 shows a process in which a person image region 2 is cut out from the person image digital image 1 and separated from the person image region 2 and its background image 3 in the person image digital image 1 taken from the front of the person. ing. Digital image 1 of a person photograph is taken by photographing the upper body of the person from the front of the person with a camera-equipped mobile phone or digital camera, or by scanning a photograph of the person taken from the front of the person with an ordinary camera (film camera) or the like. It is obtained by digitizing a necessary part of a photograph by means such as.

First, as shown in FIG. 1, by using a region dividing method, the characteristics of the person image area 2 and the characteristics of the area of the background image 3 are identified, and the person image is extracted from the digital image 1 of the person photograph. The area 2 is cut out and extracted, and the person image area 2 and the background image 3 are separated. If the background image 3 is the same color as the person and is plain, the person image area 2 can be easily extracted by an image processing technique such as edge detection of the person image. The separated image area 2 of the person is pasted on the rotating body model 4.

Next, FIG. 2 shows a generation process of the rotating body model 4 for pasting the separated human image area 2. The rotating body model 4 has a size corresponding to the size of the upper body of the person and uses a solid formed by rotating around the center line (median line) of the upper body area of the person. The human image area 2 is placed on xy orthogonal coordinates, and has a diameter equal to the width D (y) for each height in the y direction of the human image area 2 and is centered on a rotation axis parallel to the y axis. Generate as a rotating body. Since the digital image 1 of the portrait is a front face, it is basically symmetrical. For this reason, the rotating body model 4 is generated with a center line parallel to the y axis at the position in the x direction of the human image region 2 as the rotation axis. In this way, the outer shape of the rotator model 4 matches the contour of the parallel projection of the human image area 2. Normally, the rotating body model 4 is colorless and transparent.

Next, FIG. 3 shows the positional relationship among the rotating body model 4, the person image area 2, and the camera 5. The human image area 2 separated from the digital image 1 of the human photograph is pasted on the rotating body model 4. For this purpose, the rotating body model 4 and the camera 5 on the optical axis 6 of the camera 5 are attached as shown in FIG. An image area 2 of a person is placed between The optical axis 6 of the camera 5 is perpendicular to the central axis of the rotating body model 4. For convenience, the camera 5 is considered to be at infinity with respect to the rotating body model 4 and the human image area 2.

FIG. 4 shows a process of pasting the person image region 2 onto the rotating body model 4. This pasting process is based on the positional relationship shown in FIG. 3, and the image area 2 of the person is applied to the rotating body model 4 using a picture mapping technique, for example, a parallel projection mapping technique, as shown in FIG. This is done by pasting.

FIG. 5 shows a reconstructed image 7 of a person image. When the image area of the rotator model 4 viewed from the camera 5 is pasted on the background image 3 separated in FIG. 1 in the positional relationship of FIG. 3, a reconstructed image 7 of a person image is obtained. This reconstructed image 7 is almost the same image as the digital image 1 of the human photograph in FIG.

FIG. 6 shows a process in which the rotating body model 4 is rotated about the rotation axis by a specified rotation angle θ. As shown in FIG. 6, this time, the rotating body model 4 is rotated by the designated rotation angle θ when viewed from the camera 5. By this rotation, the person in the image changes direction by a specified rotation angle θ.

FIG. 7 shows a camera image obtained by pasting a lattice pattern image on the rotating body model 4 instead of the person image and rotating the rotating body model 4 at a specified rotation angle θ. It can be seen from FIG. 7 that the partial image of the planar lattice pattern is greatly deformed with the rotation of the designated rotation angle θ (0 °, 10 °, 20 °, 30 °). Therefore, by rotating the rotator model 4 by the specified rotation angle θ, the orientation of the person in the image area 2 of the person pasted on the rotator model 4 can be changed.

FIG. 8 shows a process in which the image area 2 of the person of the rotating body model 4 after the rotation of the specified rotation angle θ is pasted on the background image 3 again. As described above, the reconstructed image 7 of the person image is obtained by pasting the image area 2 of the person of the rotating body model 4 after the rotation of the specified rotation angle θ as seen from the camera 5 on the background image 3 again. . Here, in the reconstructed image 7 of FIG. 8, a parallel horizontal line portion (horizontal hatched portion) at the boundary portion between the right side of the person image area 2 and the background image 3 is displayed in the original person image area 2. Although the image area exists as a part of the image, it is an area without image data generated because the width viewed from the camera 5 is reduced by the rotation of the specified rotation angle θ of the rotating body model 4. If there is an area without image data in the reconstructed image 7 in this way, the person's reconstructed image 7 does not look good. Such inconvenience can be solved by filling a region without image data (horizontal hatched portion) with the partial image of the background image 3.

FIG. 9 shows a process of filling an area without image data (horizontal hatched portion) using the partial image of the background image 3. As described above, the presence of an area without image data in the reconstructed image 7 of the person image deteriorates the appearance of the reconstructed image. For this reason, as shown in FIG. 9, in the background image 3 after the person image area is cut out, the area where the person image area 2 was located is filled using the partial image of the background image 3. In this example, since the background image 3 is white, an area without image data (horizontal hatched portion) is filled with white.

FIG. 10 shows a reconstructed image 7 of a person image by the person image area 2 of the rotating body model 4 and the embedded background image 3. Before the person image area 2 of the rotating body model 4 after the rotation of the specified rotation angle θ of the rotating body model 4 is pasted on the background image 3, as shown in FIG. By performing an operation to fill the region with the partial image of the background image 3, the reconstructed image 7 becomes an image having no region without image data as shown in FIG. In this way, the parallel horizontal line portion (horizontal hatched portion) of the reconstructed image in FIG. 8 becomes the background image.

FIG. 11 shows a reconstructed image 7 of a human image when the designated rotation angle θ is 0 °, 10 °, 20 °, and 30 °. The reconstructed image 7 with the specified rotation angle θ = 20 ° and 30 ° is unnatural to the eye. This is because, as a first condition, the right side of the person's image area 2 appears to be missing, and as a second condition, the person's orientation appears to be tight.

FIG. 12 shows an explanation for numerically measuring the above two conditions. In FIG. 12, the center line of the image area 2 of the person is L1, and the contour line is L2. The person image region 2 is pasted on the rotating body model 4, and the center line seen from the camera 5 after rotation of the designated rotation angle θ is L3, and the right outline is L4. Further, a is the distance between the contour line L2 and the center line L3, b is the distance between the center line L3 and the contour line L4, and D is the distance between the left and right contour lines L2. These distances a, b, and D vary depending on the height, but their ratio is constant.

First, the first condition, that is, the right side of the image area 2 of the person appears to be missing. This first condition is expressed by the parameter R1 = (a + b) / D = (1 + cos θ) / 2. it can. When the specified rotation angle θ = 0 °, the parameter R1 = 1, and when the specified rotation angle θ = 90 °, the parameter R1 = 0.5. Previous experiments have shown that the parameter R1 appears to be missing when it is smaller than 0.99 to 0.98.

Next, the second condition, that is, the direction seems to be tight, can be expressed by the parameter R2 = a / (a + b) = (1−sin θ) / (1 + cos θ). When the specified rotation angle θ = 0 °, the parameter R2 = 0.5, and when the specified angle θ = 90 °, the parameter R2 = 0. It has been found in a prior experiment that the parameter R2 is about 0.4 and looks good and looks sideways.

Table 1 shows the relationship between the specified rotation angle θ and the parameters R1 and R2. It can also be seen numerically that when the specified angle θ = 10 ° to 15 ° is exceeded, the person image appears to be missing and the direction appears to be tight. For this reason, the reconstructed image having the best appearance can be obtained when the specified rotation angle θ = 10 ° to 15 °.

The person photo image processing method of the present invention can be applied to all human face photos, so if there is a photo of the front of the person, create a face photo with a different orientation and display it on a newspaper advertisement or on the Internet screen it can. Note that the picture mapping technique for pasting the person's image area 2 onto the rotator model 4 is not limited to parallel projection mapping, but uses lap mapping or the like for each characteristic surface of the rotator model 4 and there is no image data. It is also possible to synthesize the reconstructed image 7 of the person image without burying the background image 3 by eliminating the portion.

It is explanatory drawing of the process of extracting the person's image area | region 2 from the digital image 1 of a person photograph. FIG. 5 is an explanatory diagram of a generation process of the rotating body model 4. It is explanatory drawing of the positional relationship of the person's image area | region 2, the rotary body model 4, and the camera 5. FIG. It is explanatory drawing of the process which affixes the person's image area | region 2 on the rotary body model 4. FIG. It is explanatory drawing of the process which pastes the image area 2 of the person of the rotary body model 4 on the original background screen 3. FIG. It is explanatory drawing of the process in which the rotating body model 4 is rotated only by the designated rotation angle (theta). It is explanatory drawing of the camera image obtained by rotation of the designated rotation angle (theta) of the rotary body model which stuck the grid | lattice image. It is explanatory drawing of the process in which the image area 2 of the person of the rotating body model 4 after the rotation of the designated rotation angle θ is pasted on the background image 3 again. It is explanatory drawing of the process of filling up the area | region which had the person's image area 2 using the partial image of the background image 3. FIG. It is explanatory drawing of the process affixed on the background image 3 by which the image area 2 of the person of the rotary body model 4 after the rotation of the designated rotation angle (theta) was embedded by the partial image of the background image 3. FIG. It is explanatory drawing of the reconstruction image 7 for every designated rotation angle (theta). It is explanatory drawing of the outline curve and centerline of a person image after rotation of the designated rotation angle (theta).

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Digital image of person photograph 2 Image area of person 3 Background image 4 Rotating body model 4
5 Camera 6 Optical axis 7 Reconstructed image D (y) Width θ Specified rotation angles L1, L3 Center lines L2, L4 Contour lines a, b, D Distance R1, R2 Parameters

Claims (3)

In the digital image (1) of the human photograph taken from the front of the person, the human image area (2) is cut out from the digital image (1) of the human photograph to extract the human image area (2) and the human background image (3). ), The process of pasting the separated human image area (2) onto the rotating body model (4), the process of rotating the rotating body model (4) by the specified rotation angle, and the direction of the rotation change depending on the rotation. In the process of reconstructing the digital image (1) of the person photo by pasting the image area (2) of the person in the rotating body model (4) on the background image (3) separated in the above process, A method for processing a person photo image, characterized in that the orientation of the person is changed. The rotating body model (4) has a size corresponding to the size of the upper body of the person, and a solid formed by rotating around the center line of the region of the upper body of the cut out person is used as the rotating body model (4). 2. A method for processing a human photographic image according to claim 1, wherein: The background image (3) after the person image area (2) is cut out, and the area where the person image area (2) is located is filled with a partial image of the background image (3). Item 2. A method for processing a human photograph image according to Item 1.