KR20100012948A - Image dressing system for composing goods and user image in internet - Google Patents

Abstract

PURPOSE: An image dressing system for combining a product image of an internet shopping mall with a user image is provided to determine whether a desired product is proper by combination of a product image, thereby selecting a proper product. CONSTITUTION: A web service unit(100) provides an interface screen for providing a product purchase service on the internet. An image extracting unit(120) extracts a product image and a combination target area image of a user image. An image combining unit(130) combines the extracted product image with a combination target area image of the user. The image combining unit transmits a combination result to a user terminal(200) through the web service unit. A storage unit(140) stores interface screen information, product information and product image information.

Description

Image dressing system for compositing product images from internet shopping malls into user images {IMAGE DRESSING SYSTEM FOR COMPOSING GOODS AND USER IMAGE IN INTERNET}

According to the present invention, a product image of an internet shopping mall which extracts a synthesis target region image to be extracted from a user's image and then synthesizes a product image such as clothing and personal ornaments in the internet shopping mall is synthesized into a synthesis target image. An image dressing system.

Recently, data transmission speed has increased dramatically due to the spread of the Internet leased line through the optical cable, and the number of Internet users has increased exponentially due to the improvement of the Internet environment due to the development of wireless Internet technology. Sites that provide various services are emerging, the most representative of which is the Internet shopping mall.

Shopping malls are not only existing on the Internet, but are concepts that already exist in the real world. Indeed, there are many shopping malls such as Walmart and E-mart, and the Internet shopping mall is the internet shopping mall. In other words, a cyber shopping mall is a form in which a purchaser makes a purchase (credit card, electronic money, cash, etc.) through a website established on the Internet without directly visiting a merchandise distributor when a buyer wants to purchase the product. These Internet shopping malls are often referred to as Electronic Commerce (EC) companies.

As described above, the Internet shopping malls have a sales strategy that emphasizes their distinctiveness from other sites in order to provide a unique service type, and make purchases due to convenience of access, development of a payment method, and development of a delivery system. In the shopping-related field, it has become the equivalent of offline shopping.

However, the above-described Internet shopping malls of the related art have a catalog-level configuration shown to the buyer when the purchase process is simple and the buyer selects the product while the purchase process is simple and the price comparison function is biased. In other words, in the existing cyber shopping mall, a product listed according to the classification criteria of a shopping mall system or a buyer searches through a search engine, searches an image, document information, etc. of the product according to the search result, and then the buyer selects a desired product. It is focused on purchasing items such as methods.

In the above-described Internet shopping mall of the related art, the method proposed for product selection is limited to providing an image view-oriented service as a method for selecting a product of the user, and thus, the Internet shopping mall selling clothes and personal ornaments. In this case, it is difficult for a user to determine whether the product is actually suitable for himself or herself using only the image of the internet shopping mall.

In order to overcome this problem, the conventional Internet shopping malls provide a service for indirectly checking a user's wearing state by purchasing a user's product using a flash game such as dressing a doll, but this is a fixed avatar on the Internet. It's just a simple way to put on your jewelry. In addition, since the user does not use the actual picture above, the user may not be able to determine whether the product is suitable for his or her body shape.

 Therefore, the present invention is to solve the above-mentioned problems of the prior art, when the user selects a product by accessing a shopping mall selling medical, jewelry, etc., the composition target area such as the user's own body area from the image of the user The product image of the Internet shopping mall that allows the user to provide a product selection service such as the user can select the product after actually wearing it offline by extracting the image of the product and composing it as if the product is worn. It is an object to provide an image dressing system for compositing.

An image dressing system (hereinafter referred to as an "image dressing system") for compositing a product image of an Internet shopping mall of the present invention to a user image for achieving the above object is an interface screen for providing a product purchase service on the Internet. Providing a web service unit; An image extraction unit for extracting a product image selected by the user connected through the web service unit and a composition target region image from the user image image; An image synthesizing unit configured to synthesize the merchandise image extracted by the image extracting unit to the synthesis target region image of the user and transmit the commodity image to the user terminal through the web service unit; And a storage unit for storing interface screen information, product information, and product image information of the web service unit.

The web service unit may be implemented as a server system program for providing an Internet service such as IIS or Apache server.

The interface provided by the web service unit may include: a product information output interface screen for outputting product information to a user terminal connected through the Internet, a payment interface screen for performing payment, and the user to extract a region to be synthesized from his / her own image; And a composite image output interface screen for synthesizing a user image extraction interface for interacting with each other, and a synthesis target region image among the user selected product image and the user image image.

The user image extraction interface of the above-described configuration is characterized in that configured to receive the user's body information.

The image extraction unit extracts an image of the composition target region by an edge extraction method, and the edge extraction method includes CANNY, pseudo-operator, differential operator, edge optimal threshold detection method, noise removal, SNAKE, and the like. The outline extraction algorithm, the method of using the shape information of the object that requires the object based on the geometric characteristics, the method of determining the accuracy of the object through the captain learning process of the standardized standard object, the method of using the color information of the image At least one method of using a histogram and a method using a probability distribution of color information may be applied.

When the image extracting unit extracts the synthesis target image from the user's image, the image extracting unit extracts a region designated by the user as the synthesis target region through the user image extraction interface provided through the web service unit. In this case, the edge extraction method, a skin color detection algorithm, and a clustering algorithm such as KMEANS may be applied to the user's image extraction.

The image synthesizing unit adjusts the size of the merchandise image according to the user input body size information to synthesize the merchandise image with the user's synthesis target region image extracted by the image extracting unit. In this case, morphing and warping operations are applied to the image synthesis so that the synthesized image is naturally synthesized.

An image dressing system having the above-described configuration is generally implemented as a server computer having a central processing unit, an external input device, a display device, a communication device, a storage device, an output device, and the like, configured to provide an Internet service on the Internet. It can be obvious.

The image dressing system of the present invention having the above-described configuration, when a user selects a product, extracts the corresponding product image, and after the user extracts the image of the composition target region from the user image image, such as the user photo through the web service unit By synthesizing the corresponding merchandise selected by the user to the extracted synthesis target image, the user can determine whether the product is suitable for him by generating and outputting the same image as the user wears.

That is, the present invention allows a real user to select a product that is more suitable for him by virtually trying to wear the corresponding product using a product image (clothes, shoes, a hat, etc.) in his photograph on the Internet.

In addition, the present invention described above may be implemented in a variety of forms, such as changing the actual style of their hair on the Internet using their photos in addition to the shopping mall.

In addition, it is apparent that the present invention described above can be applied to not only two-dimensional images but also three-dimensional image images.

The image dressing system of the present invention that performs the above-described configuration and operation is more suitable to the user by virtually wearing the product by using the product image (clothes, shoes, hats, etc.) on the user's own image on the Internet. It is possible to prevent the user from making a wrong purchase, thereby preventing the waste of time and the waste of time caused by the increase in the shipping cost of the user and the Internet shopping mall operator. .

In addition, the present invention can expand the use of the user by ensuring the correct purchase performance of the user, thereby providing an effect that the Internet shopping mall provider can increase the sales.

Hereinafter, the present invention will be described in more detail with reference to the accompanying drawings.

1 is a block diagram of an image dressing system according to an embodiment of the present invention.

As shown in FIG. 1, the image dressing system of the present invention is connected to the image dressing server 100 for the sale of goods on the Internet through a communication network 300 enabling the Internet communication, and the user is connected to a mobile communication terminal ( 210, the user terminal 200, such as a personal computer 220, is connected to the image dressing server 100, and outputs a composite image of a user image and a product image, and then selects and purchases a product.

In the above-described configuration, the image dressing server 100 may include: a web service unit 110 for extracting a synthesis target region image from a user image registered by a user and then synthesizing and outputting the product image selected by the user; An image extraction unit 120 for extracting a product image selected by the user connected through the web service unit 110 and a composition target region image from the user image image; An image synthesizing unit 130 for synthesizing the merchandise image extracted by the image extracting unit 120 to the synthesis target region image of the user and transmitting the commodity image to the user terminal through the web service unit; A storage unit for storing interface screen information, product information, product image information, member information, buyer information, delivery address information, various driving program information, etc. of the web service unit; and a communication unit for allowing internet access although not shown in the drawing; The server system includes an input unit configured to receive a user control command, a display unit for displaying driving state information, an output unit for outputting internal information, and the like.

The image dressing server 100 according to an embodiment of the present invention having the above-described configuration has a feature in that it targets an actual photographic image, not a simple model or a processed image. Each configuration of the image dressing server 100 having such a feature will be described in more detail.

The web service unit 110 of the configuration of the image dressing server 100 is implemented as a server system program for providing an Internet service such as IIS, Apache server, etc. which is implemented in a server system to provide an Internet service on the Internet.

And the web service 110 is configured as described above the web service is output from the Internet browser, such as Internet Explorer, Netscape running on the user terminal 200 of the interface to enable the user to interact through the Internet Control the drive.

Here, the interface provided by the web service unit 110 may include a product information output interface screen for outputting product information to a user terminal connected through the Internet, a payment interface screen for performing payment, and the user synthesized from his image image. A user image extraction interface for interacting to extract the target region, and a composite image output interface screen for synthesizing the synthesis target region image among the user-selected product image and the user image image.

In addition, the user image extraction interface of the above-described configuration is configured to receive the user's body information so that the image synthesis unit 130 can modify the size of the product image to be synthesized according to the user body information.

Next, the image extracting unit 120 extracts a synthesis target image from the user's image and performs a function of extracting a product image.

At this time, the image extraction unit 120 extracts an image of the composition target region by an edge extraction method, etc. The edge extraction method is CANNY, pseudo-operator, differential operator, edge optimal threshold detection method, noise removal. Outline extraction algorithm, SNAKE, etc., how to use the shape information of the object that needs the object based on the geometric characteristics, how to determine the accuracy of the object through the captain learning process of standardized standard object, how to use the color information of the image As a method, one or more of a method using a histogram of color information and a method using a probability distribution of color information may be applied.

When the image extraction unit 120 extracts the synthesis target image from the user's image, the image extraction unit 120 extracts a region designated by the user through the user image extraction interface provided through the web service unit. At this time, the image extraction of the user applies the edge extraction method, a skin color detection algorithm, and a clustering algorithm such as KMEANS.

The image extraction method of the above-described image extraction unit 120 will be described in more detail as follows.

The image extractor 120 extracts a commodity image and a composition target region image to be synthesized from the user image uploaded by the user. That is, the image extractor 120 extracts the object in the actual photographic image to perform effective image synthesis between the images.

Currently, among the various approaches for extracting an object in a single photorealistic image having various backgrounds, there is a method of extracting an edge in a method of extracting an object from an image. Edges are the easiest way to see the relationship between image objects in an image. There are many ways to extract edges using CANNY, similarity operators, and derivative operators. Images extracted by the edge extraction method may have different threshold values depending on the image, and it is necessary to find an optimal threshold value. In addition, since the amount of noise generated varies depending on the image, an image filter for removing noise is also required.

In order to extract an object through the edge information, it is necessary to be able to extract the outer edge of the object. Therefore, an outline tracking algorithm such as SNAKE, which can trace the outline of the object, is applied.

The method using shape information of object based on geometric features is also applied to better object extraction. The method using the shape information of the object based on the above-described geometrical features makes it possible to determine whether the object extracted based on the edge has the required feature of the object through the shape information. In addition to determining the feature of the object by geometric features, the accuracy of the object can be determined through the selected learning process.

The image extraction by the above-described image extraction unit 110 is also applied to the method using the color information of the image itself in addition to the above-described edge, geometric shape information, learning process, such method using a histogram of the color information For example, the color information includes a threshold value, a chroma key method commonly used for broadcasting, a probability distribution method of color information, and the like.

In the case of a product image of an internet shopping mall, the background is usually of a single color. Therefore, analyzing a background of a single color may make it easier to extract a product image in an internet shopping mall than a complex photorealistic image.

In addition, in the case of a product image of a single background, since a relatively simple edge exists for each object, the product image object can be extracted by a combination of edges, geometric shape information, learning process, and color information. In the case of photorealistic pictures with complex backgrounds used by users, this information alone may be difficult to accurately extract objects. Therefore, this is solved by allowing the user's interaction to correct the error. To this end, the web service unit 110 provides a user image extraction interface that allows the user to select a composition target region from the user image uploaded by the user in the user terminal. The user may select a region to be synthesized among user images registered by the user on the user image extraction interface output to the Internet browser of the user terminal 200, and input the user's body information together to select an appropriate size of the product image. It is composed.

For example, the image extraction method of the image extraction unit 120 will be described in more detail as follows.

2 is a view for explaining a method of extracting a product image having a single background provided in an internet shopping mall, and shows an actual product image provided in an internet shopping mall having an image dressing server 100.

As shown in FIG. 2, an internet shopping mall image having a single background may extract an object as follows. First, the method that can operate in the case of a shopping mall image of a single background will be described. In the case of a single background, a complex background does not exist, and chroma key method and edge detection method can effectively separate objects. The chroma key method extracts an object by finding a feature color of a background and setting a threshold based on the found background feature. This is to use the following equations (1) to (4) to grasp the characteristics of the background, weight the feature color, project the color space into the space of Y, Cb, Cr, and analyze the difference with the Y space to separate the object. Can be.

In Equation 1, when the background is a green series, when RGB and Bgain are projected in the Y space, Rgain and Bgain decrease the gain by a and conversely, add the gain by b for the G region.

Y (green) = (R-a) * Rgain + (B-a) * Bgain + (G + b) * Ggain

The Y-space projected image obtained as described above may obtain the difference between Y (green) and the original Y-space region using Equations 2 to 4, and may be separated by subjecting the normalization process to a value of 255. N is the weighted image for the object and c is the gain between Y (green) and Y.

N = c * (Y (green) -Y)

N (i, j)> threshold th (threshold)-> N (I, j) = 255

N (i, j) <threshold th-threshold N> (I, j) = 0

In the above formula, the green background is taken as an example, and information of Y (background color) can be used according to the background color. (See "A Study on the Development of Handset Contents of Human Background Synthesis Using WIPI."

Next, FIG. 3 is a diagram illustrating an object extraction method under a green and blue background. When object extraction is difficult only by separation of a background color, edge information is added to the object separation.

Next, FIG. 4 is a diagram illustrating an image of edges extracted by applying a noise filter and edge extraction to the image of the Internet shopping mall of FIG. 2.

After edge extraction, segmentation is performed through edge search. The object is extracted through the segmentation process. In addition to segmenting in the direction of the arrow, you can apply an outline tracking algorithm such as snake to separate the objects.

Edge extraction methods include the following.

Pseudo-operator technique: The simplest and fastest edge extractor is used to determine the maximum value subtracted from the center pixel of a 3x3 block and to use the largest of the absolute values of each difference.

Difference Operator Technique: The similar operator uses eight operations, but the difference operator calculates it only four times.

Differential operator technique: Extract edges using convolution results in horizontal and vertical directions (Roberts, prewitt, sobel)

Quadratic Differential Operator Technique: Compensating for the Drawbacks of First Differential (laplacian, laplacian of Gaussian, differential of Gaussian)

Canny: Contours the image with a Gaussian mask and calculates the tilt direction through the tilt element between each pixel. The second derivative is obtained based on the slope direction, and the position of the zero value of the second derivative is classified as an edge point.

This separating object object is not only clothing but also personal ornaments, etc., as well as all the goods in the Internet shopping mall.

FIG. 5 is a diagram illustrating an extraction of a composition target region image from a user image uploaded by a user, and describes a method of extracting an object from a photo uploaded by a consumer with reference to FIG. 5.

When the user displays a picture area to which he belongs in the picture, the user's object is extracted from the displayed picture. The extraction process is as follows.

For edge information extraction, edge image is made after preprocessing and noise filtered image is obtained. In the case of a complex background, edges can be extracted even in the background area, and the skin color area is extracted by applying a skin color detection algorithm to identify a user object. Through matching the skin region and edge image, we find the center point of the edge that matches the skin region and analyze the outline using the outline extraction algorithm such as snake around the center point. In addition to extracting user image objects through the analysis of the skin area, the contour line may be analyzed after segmentation by color area using a clustering algorithm such as kmeans. In addition to the outline analysis, the object region may be classified by region labeling. In addition, it is possible to classify regions through a labeling operation using a clustering distance value using wavelets.

Through this series of processes, the object area is analyzed and the shape information of the object is analyzed to determine whether it meets the characteristics of the object. If it matches the characteristics of the object, it recognizes it as the correct object and extracts its area. The shape information of the object can be determined by using a geometric method, neural network about the object features, or through the learning process of a classifier such as svm. If the result of the judgment matches the object to be extracted, the object is separated. It is very difficult to completely separate objects within various backgrounds, allowing for interaction with the user in case of errors. User interaction refers to the way the user directly displays the object he / she wants. In addition, the method included in the product image extraction process applied earlier may also be additionally applied.

The color detection algorithm applied to user image extraction method uses empirical method, color histogram method, Single Gaussian Model (SGM), Gaussian Mixture Model (GMM), and Applicant's own developed color detection algorithm. In addition, it is also possible to investigate the distribution of various skin color and to determine the skin color area by learning the result.

The equation representing the flesh color detection algorithm applied in the above-described user image extraction method is as follows.

So far, we have explained how to separate objects when uploading your own photorealistic images. Next, we explain how to create a model through a model. First, a human body model exists and a user inputs a body size such as his height and weight in the human body model. Also, by modifying each part of the human body model, it is possible to create a model very similar to itself.

Next, the image synthesizing unit 130 adjusts the size of the product image according to the user input body size information and synthesizes the product image with the synthesis target region image of the user extracted by the image extracting unit. At this time, warping, morphing, and metamorphosis work are applied to the image synthesis, so that the synthesized image is naturally synthesized.

Before CG shape deformation technology emerges, it is also called Dissolve (commonly known as OVER LAB) by using a special make-up.If you double-print the back shot after fading in on the front shot that fades out, the front shot It gradually disappears and the back shot gradually appears.) However, this resulted in various unrealistic results during dissolve and cost a lot of money and time. But now, with the development of CG technology, all these problems have been solved, which is Warphing, Morphing, Metamorphosis. These all mean a kind of shape transformation work.

In more detail,

Warping (or sometimes referred to as shape animation) refers to a transformation from A to A 'that distorts the expression but leaves the shape intact. In other words, the texture or texture of the object is not changed, only the shape is deformed. For example, a cat smiles in a refrigerator advertisement, a cow smiles in an old advertisement, an air conditioner bulges in an air conditioner advertisement, and a newborn baby moves his mouth in line with such an example.

Morphing refers to a technique of gradually deforming from A to B a shape of an object in a completely different shape. For example, in a music video, the scene changes between races.

In addition, metamorphosis is a shape deformation of a CG object manufactured in three dimensions. In contrast, metamorphosis refers to a shape deformation of a three-dimensional object made in CG, while morphing is a shape deformation between real images taken. And while most cameras are static in morphing, Metamorphosis has the advantage of allowing the camera to move freely from any angle. However, companies often reduce metamorphosis and call it morphing.

6 is a flowchart illustrating a process of an image synthesizing process by the image dressing system of the present invention.

As shown in FIG. 6, a user who wishes to purchase a product may access the image dressing server 100 through a communication network 300 capable of internet using the user terminal 200, and then perform a product search to purchase. We choose product to say. When the user selects the desired product to purchase, the image extracting unit 120 of the image dressing server 100 searches for and extracts the product image stored in the storage unit 140 or from the product information image provided by the web service unit 110. The product image is extracted, and the image extraction method uses the methods described in the above description of the image extraction unit 110 of FIG. 1 (S10).

As described above, after the user selects the desired product to be extracted and the product image is extracted, when the user uploads his own image including the video image to be synthesized to the image dressing server 100, the image dressing server 100 The image extractor 120 extracts an image image to be synthesized from the product image from the image. In this process, the image extraction method described in the above description of the image extraction unit 120 of FIG. 1 is applied.

In addition, in the above-described process, the web service unit 110 transmits the user image extraction interface screen to the user terminal 200 so that the user can select a synthesis target region image from the user image image, and then synthesizes the image among the images uploaded by the user. It is configured to select an area. In this case, the user image extraction interface is configured to perform line drawing, color selection, and the like, and also inputs the user's body size and weight information to adjust the size of the product image to be synthesized in the user image in the image synthesis unit 130. Configured to do so.

 When the user selects the synthesis target region through the user image extraction interface screen, the image extractor 120 extracts an image of the synthesis target region from the user image image by applying the image extraction method described in the above description of FIG. S20).

As described above, after the composition target region image is extracted from the product image and the user image image, the image synthesis unit 130 adjusts the size of the product image according to the body information input by the user and then synthesizes the region of the user's composition. After compositing the corresponding product image to the image and transmits it to the user terminal through the web service unit 110 so that the user can check the wearing state of the product. FIG. 7 illustrates a commodity image and a synthesized image to be synthesized with the synthesis target image image among the user images to be synthesized by the image synthesizing unit 130 described above. In this case, clothing, ornaments, etc. are synthesized so that the user wears clothing and ornaments, and in this process, the morphing and warping operations are performed for natural synthesis as described above (S30).

1 is a block diagram of an image dressing system according to an embodiment of the present invention;

2 is a view for explaining a method of extracting a product image having a single background provided in an internet shopping mall;

3 is a view showing an object extraction method applied under a green and blue background;

4 is a diagram illustrating an image of edges extracted by applying a noise filter and edge extraction to the image of the Internet shopping mall of FIG. 2;

FIG. 5 is a diagram illustrating an extraction of a composition target region image from a user image uploaded by a user; FIG.

6 is a flowchart showing a process of an image combining process by the image dressing system of the present invention;

FIG. 7 is a diagram illustrating a merchandise image to be synthesized, a commodity image to be synthesized, and a synthesized image among the user images to be synthesized by the image synthesizing unit 130 described above.

Description of the main symbols in the drawings

100: image dressing server 110: Web service unit

120: image extraction unit 130: image synthesis unit

140: storage unit

200: user terminal

210: mobile communication terminal 220: personal computer

300: network

Claims (5)

A web service unit providing an interface screen for providing a product purchase service on the Internet; An image extraction unit for extracting a product image selected by the user connected through the web service unit and a composition target region image from the user image image; An image synthesizing unit configured to synthesize the merchandise image extracted by the image extracting unit to the synthesis target region image of the user and transmit the commodity image to the user terminal through the web service unit; And a storage unit for storing interface screen information, product information, and product image information of the web service unit. The method of claim 1, wherein the web service unit, the product information output interface screen for outputting the product information to the user terminal connected through the Internet, the user to perform the interaction so that the user can extract the composition region from the video image And a user image extraction interface, and an interface including a composite image output interface screen for synthesizing a synthesis target region image among the user selected product image and the user image image. Image dressing system to synthesize. The image dressing system of claim 2, wherein the user image extraction interface is configured to receive input of user's body information. The image dressing system of claim 4, wherein the image extracting unit extracts a region designated by the user as a composition target region through a user image extracting interface provided through a web service unit. . The Internet shopping mall of claim 3, wherein the image synthesizing unit adjusts the size of the commodity image according to the user input body size information and synthesizes the commodity image with the synthesis target region image of the user extracted by the image extracting unit. Image dressing system to combine the product image of the user image.

Images