JP3706413B2 - Image composition method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an image composition method for replacing an image of an object photographed with a monochromatic background with a new background image in the creation of an electronic catalog or the like.
[0002]
[Prior art]
In recent years, attempts have been made to digitize catalogs of a large number of products, increase the efficiency of product searches, and to make product orders online (Japanese Patent Application Laid-Open No. 7-56784 [G06F12 / 00]). Kaihei 4-247567 [G06F15 / 21]).
In creating such an electronic catalog, first, a product is photographed with a camera on a monochrome background, and a color image obtained thereby is captured by an image scanner or the like and digitized. The digitized original image is sent to an image processing circuit by a microcomputer.
[0003]
The digital original image is composed of, for example, a monochrome background portion U as shown in FIG. 10 (a) and an object portion V which is an image of a product. As shown in (b), it is replaced with a new arbitrary background image U ″.
Accordingly, by changing the background image U ″ to be newly synthesized into a natural landscape image or an artificial pattern image according to the type of product, various aesthetic effects can be given to the product.
[0004]
At the time of image composition, first, a pixel that is clearly estimated to be a background portion is selected for the digital original image of FIG. 10A, and then the next pixel that is adjacent to the selected pixel and has no significant difference in pixel value. Pixels, that is, adjacent pixels that do not exceed a predetermined threshold are sequentially selected. Here, as pixel values, for example, luminance data obtained by performing YIQ conversion on RGB three primary color signals ("0" to "255") can be employed.
As a result, the digital image is separated into an object area and a background area as shown in FIG. 11A based on the threshold value.
Therefore, the synthesized image is obtained by replacing the pixel value of the background area with the pixel value of the new background image as shown in FIG.
FIGS. 11A and 11B schematically show changes in pixel values due to changes in the one-dimensional direction of pixel positions.
[0005]
[Problems to be solved by the invention]
However, in the digital original image, the boundary between the background region and the object region is blunted as shown in FIG. 11A due to quantization error at the time of digitization by an image scanner or the like, and several pixels straddling the background region to the object region. The pixel value continuously changes within the region.
As a result, when the background region is replaced with a new background image as shown in FIG. 11B, the change in the pixel value of the composite image becomes discontinuous between the new background image and the object image, and the quantization of the object image In the ring-shaped region R blunted by the error, a mixed color with the original background color remains, resulting in an unnatural composite image in which the periphery of the object is bordered with the mixed color as shown in FIG. was there.
[0006]
The bordering problem can be alleviated to some extent by setting a high threshold value. However, the pixel value of the object region varies greatly depending on the location, and the threshold value has to be set lower than the pixel value of the pixel closest to the background region among the pixels of the object region. The value cannot be set sufficiently high, and it is difficult to eliminate the bordering problem.
An object of the present invention is an image in which an unnatural boundary region is not generated between an object image and a background image in an image composition process in which a background portion of a digital original image is replaced with an arbitrary digital background image. It is to provide a synthesis method.
[0007]
[Means for Solving the Problems]
The image composition method according to the present invention includes a first step of separating a digital original image into a background area and an object area by a predetermined threshold, and setting a boundary portion between the background area and the object area, Is divided into three regions, a background part, an object part, and a boundary part, the background part is replaced with a digital background image, the original part is maintained for the object part, and the object part is And a third step of generating a connection image in which pixel values are smoothly changed from the original image to the digital background image, and synthesizing these three images.
Specifically, in the third step, for each pixel constituting the boundary portion of the digital original image, the pixel value of the pixel is the pixel value of the object portion pixel closest to the pixel and the background of the digital original image After calculating the mixing ratio when it is a mixed value with the pixel value of the part, for each pixel that should constitute the connection image, the pixel value of the object part closest to the pixel and the digital background image The pixel value of the pixel closest to the pixel is mixed at the mixing ratio at the same pixel position to calculate the pixel value of the pixel.
[0008]
In the image composition method, the threshold value to be set in the first step may be an arbitrary value between the pixel value of the background portion and the pixel value of the object portion of the digital original image. It is not necessary to set a delicate value. The boundary portion set in the second step is a width region including a blunt portion between the object image and the background image that has been caused by the quantization error in the digital original image regardless of the threshold value. Set to In the third step, instead of the blunted part, a connection image is generated that smoothly changes the pixel value from the original image of the object part to the digital background image. Therefore, by synthesizing the digital background image of the background portion, the original image of the object portion, and the connection image of the boundary portion on one screen, a composite image in which pixel values smoothly change from the object portion to the background portion can be obtained. .
In the image composition method, an appropriate pixel value is set for each pixel that should form a connection image.
That is, since the pixel value continuously changes at the boundary portion of the digital original image, the pixel value of each pixel is the pixel value of the object portion closest to the pixel and the pixel value of the background portion. Can be considered as a value that is mixed at a ratio according to the pixel position. By mixing the pixel value of the portion and the pixel value of the digital background image, the pixel value of each pixel constituting the connection image can be calculated as an appropriate value.
[0009]
Further, specifically, in the second step, the raw digital image, and binarizes the first pixel value of the background area (for example, "255") and the second pixel value of the object area (for example, "0") Thereafter, the binary image is subjected to a smoothing process, and a blunt area between the background area and the object area is set as a boundary part, and the remaining areas are set as a background part and an object part. According to the specific method, a boundary portion having a width (about several pixels) corresponding to the width of the unit region for the smoothing process is formed between the background portion and the object portion.
[0012]
【The invention's effect】
According to the image composition method of the present invention, the original image of the object part and the digital background image of the background part are connected with a smooth change in pixel value by the connection image of the boundary part. An unnatural boundary region is not generated between the two images, and a natural composite image is obtained.
[0013]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, embodiments of the present invention will be specifically described with reference to the drawings.
FIG. 1 shows an image composition procedure according to the present invention, and FIGS. 2A to 2D illustrate a process in which a composite image is created from a digital original image by the procedure.
[0014]
When superimposing a new arbitrary background image U ′ on a background of a full-color digital original image composed of a single-color background portion U and an object portion V, which is a product image, as shown in FIG. The following image composition process is executed.
The digital background image and the digital background image of the new background image to be synthesized are each composed of RGB three primary color signals ("0" to "255"), and are used for image separation based on thresholds described later. As the “pixel value” used, for example, luminance data Y obtained by performing YIQ conversion on the three primary color signals is employed.
[0015]
First, in step S1 of FIG. 1, a threshold value is set as a reference for separating the digital original image into a background area and an object area. Here, as shown in FIG. 3, the threshold value may be set as an arbitrary value between the pixel value of the background portion of the digital original image and the pixel value of the object portion.
In step S2 of FIG. 1, a pixel having a pixel value lower than the threshold is set as a background portion, a pixel having a pixel value higher than the threshold is set as an object portion, and the digital original image is set as a background portion and an object portion. To separate.
[0016]
The specific procedure for image separation based on the threshold is as follows. First, a pixel that is clearly estimated to be a background portion is selected, and then the next pixel that is adjacent to the selected pixel and has no significant difference in pixel value, that is, a predetermined pixel. Neighboring pixels that do not exceed the threshold are sequentially selected. As a result, the background portion is separated from the digital original image, and the remaining region becomes the object portion (see FIG. 3).
When a hollow area exists in the center of the object image and the hollow area is also replaced with a new background image, it is necessary to select and separate the hollow area as a background portion.
[0017]
Next, in step S3 of FIG. 1, the digital original image is binarized by setting the separated background portion to “255” and the object portion to “0”. As a result, a binary image as shown in FIG. 2B is obtained.
In step S4, the binary image is smoothed by a Gaussian filter having a radius of 4 pixels, for example, so that the boundary between the background portion and the object portion is blunted. As shown in FIG. In between, the boundary part where both are mixed is formed.
Here, the boundary portion is set to a width region (about several pixels) that completely includes the blunt portion between the object and the background of the digital original image indicated by a broken line in FIG.
[0018]
Then, the entire area of the digital original image is divided into three areas, that is, the boundary portion W3, the background portion W1 on both sides thereof, and the object portion W2 as shown in FIG. 2C. In step S5 in FIG. The digital original image is ternarized by setting the pixel value of the portion W1 to “255”, the pixel value of the object portion W2 to “0”, and the pixel value of the boundary portion W3 to an intermediate value thereof, for example, “128”.
[0019]
Thereafter, the following processing is executed on three ternary image areas (background part W1, object part W2, and boundary part W3).
For the background portion W1, the original digital primary image (three primary color signals) shown in FIG. 2A is replaced with a new digital background image (three primary color signals) in step S6 of FIG. For the object portion W2, the original digital original image (three primary color signals) shown in FIG. 2A is maintained as it is in step S9 of FIG.
For the boundary portion W3, in steps S7 and S8, in order to smoothly change the pixel values, that is, the three primary color signals (R, G, B) from the object portion image to the new background image as shown in FIG. A connection image is generated.
[0020]
Here, the connection image generation method in steps S7 and S8 will be described in detail. The processes in steps S7 and S8 are executed for each of the three primary color signals (R, G, B).
Since the pixel value of the boundary portion from the object portion to the background portion continuously changes in the digital original image as shown in FIG. 7, the pixel value of one point P having the boundary portion is the object portion closest to the pixel. Can be considered as a value obtained by mixing the pixel value of the pixel and the pixel value of the background portion at a ratio corresponding to the pixel position.
[0021]
That is, when the pixel value of the object portion is represented by O, the pixel value of the background portion is represented by B, and the pixel value of the one point is represented by P, the mixing ratio X is represented by the following formula 1.
[Expression 1]
X = (P−B) / (O−B)
[0022]
Similarly, in the connection image that smoothly connects the object image and the new background image as shown in FIG. 8, in order to continuously change the pixel value of the connection image, the pixel value of one point P ′ is: The pixel value of the object part closest to the pixel and the pixel value of the new background part closest to the pixel may be set to a value that is mixed at a ratio according to the pixel position.
[0023]
That is, when the pixel value of the new background portion is represented by N and the pixel value of the one point is represented by P ′, the mixing ratio X ′ is represented by the following formula 2.
[Expression 2]
X '= (P'-N) / (ON)
[0024]
Since it is appropriate to set the mixing ratio X of the formula 1 and the mixing ratio X ′ of the formula 2 to the same value, the following formula 3 is obtained from these relationships.
[Equation 3]
P ′ = {N (OP) + O (P−B)} / (OB)
[0025]
Therefore, a new pixel value can be calculated for all the pixels of the boundary image by the above equation (3).
However, the pixel value O of the object portion is the pixel value of the object portion closest to the pixel for which the pixel value is to be calculated, but the pixel value differs depending on the position of the target pixel, so that it is separately determined for all target pixels. It is necessary to calculate.
[0026]
Therefore, as shown in FIG. 5, the boundary portion W3 is divided into predetermined minute areas (for example, about several pixels), and the least square method is applied to the pixel distribution of the ternary image (pixel value “128”) in each minute area. After approximating the pixel distribution with a straight line L, a perpendicular line M is drawn with respect to the straight line L. Then, a point where the perpendicular M intersects the boundary line with the object part W2 is obtained, and the pixel value of the digital original image at the intersection is estimated as the true pixel value O of the object part (step of FIG. 1). S7).
As a result, when the boundary line between the object part W2 and the boundary part W3 is horizontal as shown in FIG. 6A, a vertical perpendicular M is drawn, and the point O closest to the point P is obtained. When the boundary line is vertical as shown in FIG. 6B, a horizontal perpendicular line M is drawn, and the point O closest to the point P is obtained. When the boundary line is inclined as shown in FIG. 6C, the inclined perpendicular line M is drawn, and the point O closest to the point P is obtained.
[0027]
In this way, after estimating the true pixel value O of the object portion, each pixel value P ′ of the boundary image is calculated by the equation 3 (step S8 in FIG. 1). By executing this calculation process for all the pixels in the boundary portion, a boundary image is generated.
Finally, in step S10, the background image obtained in step S6, the boundary image obtained in steps S7 and S8, and the object image obtained in step S9 are converted into the three primary color signals R, G, and B. Each of these is synthesized, and finally a full-color digital synthesized image composed of a new digital background image U ′ and an original image U ′ of the object part is obtained as shown in FIG.
[0028]
In the composite image, the original image of the object portion and the digital background image of the background portion are smoothly connected by the connection image as shown in FIG. 9, and the connection image has an appropriate ratio between the object image and the new background image. Therefore, the boundary between the two images is recognized as a boundary line that does not cause a sense of incongruity.
Further, according to the procedure shown in FIG. 1, since image composition can be performed by simple arithmetic processing, rapid composition processing is possible and hardware implementation is also easy.
[0029]
The above description of the embodiment is for explaining the present invention, and should not be construed as limiting the invention described in the claims or reducing the scope thereof. The configuration of each part of the present invention is not limited to the above-described embodiment, and various modifications can be made within the technical scope described in the claims.
[Brief description of the drawings]
FIG. 1 is a flowchart showing a specific procedure of an image composition method according to the present invention.
FIG. 2 is a diagram illustrating processes (a) to (d) in which an image is synthesized by the procedure.
FIG. 3 is an explanatory diagram of image separation by binarization.
FIG. 4 is an explanatory diagram of image division by smoothing processing and ternarization.
FIG. 5 is a diagram for explaining the principle of a true pixel value estimation method by a least square method.
FIG. 6 is a diagram showing a state in which true pixel values are estimated by the estimation method for three types of image patterns (a), (b), and (c).
FIG. 7 is a diagram illustrating a mixing ratio of pixel values at a boundary portion in a digital original image.
FIG. 8 is a diagram illustrating a mixing ratio of pixel values at a boundary portion in a composite image.
FIG. 9 is a diagram illustrating a distribution of pixel values in a composite image.
FIG. 10 is a diagram showing a digital original image (a) and a conventional composite image (b).
FIG. 11 is a diagram illustrating a pixel value distribution (a) in a digital image and a pixel value distribution (b) in a conventional composite image.
[Explanation of symbols]
U background part of digital original image V object part of digital original image W1 background part of ternary image W2 object part of ternary image W3 boundary part of ternary image U ′ background part of composite image V ′ of composite image Object part

Claims (2)

A method of synthesizing a digital original image obtained by digitizing an image of an object photographed with a single color background and an arbitrary digital background image to be replaced with a background portion of the digital original image,
A first step of separating the digital original image into a background area and an object area by a predetermined threshold;
A second step of setting a boundary portion between both the background region and the object region, and dividing the digital original image into three regions of the background portion, the object portion, and the boundary portion;
The background portion is replaced with a digital background image, the object portion is maintained as the original image, and the boundary portion is generated as a connection image that smoothly changes pixel values from the object portion original image to the digital background image. A third step of combining the three images ,
In the third step, for each pixel constituting the boundary portion of the digital original image, the pixel value of the pixel is the pixel value of the object portion pixel closest to the pixel and the pixel value of the background portion of the digital original image. After calculating the mixing ratio when it is a mixed value, for each pixel that should form the connection image, the pixel value of the object part closest to the pixel and the part of the digital background image closest to the pixel An image composition method for calculating the pixel value of the pixel by mixing the pixel value of the pixel with the mixing ratio at the same pixel position. In the second step, after the digital original image is binarized into the first pixel value of the background area and the second pixel value of the object area, the binary image is subjected to a smoothing process, and the digital image is processed between the background area and the object area. The image composition method according to claim 1, wherein the blunt area is a boundary part, and the remaining area is a background part and an object part.

Images