WO2019023965A1 - 一种基于智能终端的试衣方法及试衣系统 - Google Patents

一种基于智能终端的试衣方法及试衣系统 Download PDF

Info

Publication number
WO2019023965A1
WO2019023965A1 PCT/CN2017/095582 CN2017095582W WO2019023965A1 WO 2019023965 A1 WO2019023965 A1 WO 2019023965A1 CN 2017095582 W CN2017095582 W CN 2017095582W WO 2019023965 A1 WO2019023965 A1 WO 2019023965A1
Authority
WO
WIPO (PCT)
Prior art keywords
picture
clothing
portrait
smart terminal
camera
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/CN2017/095582
Other languages
English (en)
French (fr)
Inventor
张龙澎
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shenzhen Transsion Communication Co Ltd
Original Assignee
Shenzhen Transsion Communication Co Ltd
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Shenzhen Transsion Communication Co Ltd filed Critical Shenzhen Transsion Communication Co Ltd
Priority to PCT/CN2017/095582 priority Critical patent/WO2019023965A1/zh
Publication of WO2019023965A1 publication Critical patent/WO2019023965A1/zh
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T11/00—Two-dimensional [2D] image generation

Definitions

  • the present invention relates to the field of intelligent terminal technologies, and in particular, to a fitting method and a fitting system based on an intelligent terminal.
  • the invention intelligently replaces the photos of the same clothes provided by the corresponding brand official website according to the photos provided by the user, and gives the front, back and side fitting effect pictures to better show the fitting effect, and simulates the real fitting clothes as much as possible. And can give professional evaluation according to the matching degree of the user's skin color, body and other information with the fitting clothes, and the smart recommendation matches the clothing photos with higher degree of matching.
  • an object of the present invention is to provide a fitting method and a fitting system based on a smart terminal, which can realize virtual fitting in various situations and provide an optimal virtual fitting solution.
  • the invention discloses a fitting method based on a smart terminal, comprising the steps of: acquiring a portrait picture including a portrait element; acquiring a target clothing picture including a clothing element; and synthesizing the portrait picture and the target clothing picture to form a composite a picture, such that the portrait element in the portrait picture is dressed with the clothing element in the target clothing picture; the composite picture is displayed.
  • the method for acquiring a portrait image including a portrait element comprises: acquiring a portrait image by a camera photographing, or acquiring a portrait image in an application or a gallery in the smart terminal; when the method of acquiring a portrait image is acquired by taking a photo by a camera,
  • the steps include: acquiring camera rights in the smart terminal or establishing a camera module in the smart terminal; reading a portrait picture captured by the camera; and obtaining a portrait image by acquiring the portrait image in the application or the gallery in the smart terminal
  • the method includes: obtaining an access right of an application or a gallery in the smart terminal; and reading a portrait image in an application or a gallery in the smart terminal.
  • the method for acquiring a target clothing picture including a clothing element includes: acquiring a clothing picture by a camera photographing, or acquiring an application picture in the smart terminal or a clothing picture in the library; and when acquiring the clothing picture by taking a photo by the camera
  • the steps include: acquiring camera rights in the smart terminal or establishing a camera module in the smart terminal; reading a picture of the clothing taken by the camera; and obtaining the clothing picture by acquiring the application in the smart terminal or the clothing in the gallery
  • the steps include: obtaining an access right of an application or a gallery in the smart terminal; and reading a picture of the clothing in the application or the gallery in the smart terminal.
  • the portrait picture is combined with the target clothing picture to form a composite picture
  • the step of wearing the portrait element in the portrait picture with the clothing element in the target clothing picture comprises: capturing the portrait picture a portion of the portrait element that is not covered by the garment; reconstructing a contour of the garment covering portion within the portrait element according to the uncovered garment portion; covering the laundry element on the garment covering portion.
  • the method further comprises: analyzing a degree of matching between the portrait element and the clothing element in the composite picture to form a first matching degree parameter; based on the first matching degree parameter Pushing a clothing picture having a second matching degree parameter, wherein the second matching degree parameter is greater than the first matching degree parameter.
  • the invention also discloses a fitting system based on a smart terminal, comprising: a portrait picture acquisition module, acquiring a portrait picture containing a portrait element; a clothing picture acquisition module, acquiring a target clothing picture containing the clothing element; and a synthesis module,
  • the portrait picture is combined with the target clothing picture to form a composite picture, such that the portrait element in the portrait picture is dressed with the clothing element in the target clothing picture; and the display module displays the composite picture.
  • the portrait picture acquisition module comprises: a camera establishing unit, calling a smart terminal camera to capture and fix a picture; a photo reading unit reading a portrait picture in the camera; or the portrait picture obtaining module comprises: reading permission acquisition The unit sends a permission request to the smart terminal to obtain an application program and a library read permission in the smart terminal, and a photo reading unit that reads a portrait image in the application or the gallery.
  • the clothing picture obtaining module comprises: a camera establishing unit, calling the smart terminal camera to capture and fix the picture; a photo reading unit, reading the clothing picture in the camera; or the clothing picture obtaining module comprises: reading permission acquisition The unit sends a permission request to the smart terminal to obtain an application program and a library read permission in the smart terminal; and a photo reading unit that reads a clothing picture in the application or the gallery.
  • the synthesizing module includes: a portrait element capturing unit that captures an uncovered portion of the portrait element in the portrait image; a silhouette contour reconstruction unit that reconstructs the portrait element according to the uncovered clothing portion
  • the inner garment covers a contour of the portion; the covering unit covers the laundry element on the garment covering portion.
  • the system further includes: a matching degree analysis module, analyzing a matching degree between the portrait element and the clothing element in the composite picture to form a first matching degree parameter; and a clothing picture pushing module, based on The first matching degree parameter pushes a clothing picture having a second matching degree parameter, wherein the second matching degree parameter is greater than the first matching degree parameter.
  • a matching degree analysis module analyzing a matching degree between the portrait element and the clothing element in the composite picture to form a first matching degree parameter
  • a clothing picture pushing module based on The first matching degree parameter pushes a clothing picture having a second matching degree parameter, wherein the second matching degree parameter is greater than the first matching degree parameter.
  • FIG. 1 is a schematic flow chart of a fitting method in accordance with a preferred embodiment of the present invention.
  • FIG. 2 is a flow chart showing a method for obtaining a portrait picture in a fitting method according to a preferred embodiment of the present invention
  • FIG. 3 is a schematic flow chart of a method for acquiring a portrait picture in a fitting method according to another preferred embodiment of the present invention.
  • FIG. 4 is a schematic flow chart of a method for obtaining a target garment according to a fitting method according to a preferred embodiment of the present invention
  • FIG. 5 is a schematic flow chart of another method for obtaining a target laundry according to another preferred embodiment of the present invention.
  • FIG. 6 is a flow chart showing a method for generating a composite picture of a fitting method in accordance with a preferred embodiment of the present invention
  • FIG. 7 is a schematic flow chart of a fitting method according to another preferred embodiment of the present invention.
  • Figure 8 is a schematic view showing the system structure of a fitting system in accordance with a preferred embodiment of the present invention.
  • Figure 9 is a block diagram showing the structure of a portrait picture acquisition module of a fitting system in accordance with a preferred embodiment of the present invention.
  • FIG. 10 is a schematic structural view of a portrait picture acquisition module of a fitting system in accordance with another preferred embodiment of the present invention.
  • FIG 11 is a block diagram showing the structure of a clothing picture acquisition module of a fitting system in accordance with a preferred embodiment of the present invention.
  • Figure 12 is a block diagram showing the structure of a clothing picture acquisition module of a fitting system in accordance with another preferred embodiment of the present invention.
  • Figure 13 is a block diagram showing the construction of a composite module of a fitting system in accordance with a preferred embodiment of the present invention.
  • Figure 14 is a schematic view showing the structure of a fitting system in accordance with another preferred embodiment of the present invention.
  • the smart terminal can be implemented in various forms.
  • the terminal described in the present invention may include a mobile terminal such as a mobile phone, a smart phone, a notebook computer, a PDA (Personal Digital Assistant), a PAD (Tablet), a PMP (Portable Multimedia Player), a navigation device, and the like, and such as Fixed terminal for digital TV, desktop computer, etc.
  • a mobile terminal such as a mobile phone, a smart phone, a notebook computer, a PDA (Personal Digital Assistant), a PAD (Tablet), a PMP (Portable Multimedia Player), a navigation device, and the like, and such as Fixed terminal for digital TV, desktop computer, etc.
  • PDA Personal Digital Assistant
  • PAD Tablet
  • PMP Portable Multimedia Player
  • FIG. 1 is a schematic flowchart of a method for fitting a smart terminal based on a preferred embodiment of the present invention.
  • the fitting method specifically includes the following steps:
  • the portrait element refers to the main body structure of the human body: the head, the limbs, the trunk, the feet, etc., wherein the upper limb and the upper torso constitute The upper body, the lower limbs and the lower torso constitute the lower body, and the portrait elements must correspond to the target clothing.
  • the portrait image may be a half-length photograph containing only the head and the upper body; the target clothing is a dress, a skirt or a pants.
  • the portrait picture must contain the body elements of the head, upper body and lower body; when the target clothing is shoes, the portrait picture must contain the body element of the foot.
  • the target clothing is also preferably included when the shoes are shoes. Head, upper body and lower body. Remind the user to upload the corresponding picture when the user picture does not meet the above requirements.
  • the target clothing picture including the clothing element needs to be obtained, and the quality of the obtained target clothing picture directly affects the effect of the virtual fitting, and the quality of the target clothing picture here includes not only the resolution and the pixel size of the picture. It also includes the picture showing the characteristics of the clothing. Therefore, the target clothing picture must contain the complete clothing element, and clearly and completely reveal the characteristics of the clothing. Therefore, the target clothing picture is not only one piece, and the characteristics of the clothing must be exhibited from various directions, including at least the front view and the rear view, which are achieved through the plan view. The three-dimensional effect ensures a better try-on effect.
  • S300 synthesize the portrait picture and the target clothing picture to form a composite picture, so that the portrait element in the portrait picture is dressed with the clothing element in the target clothing picture.
  • the portrait picture and the target clothing picture are combined such that the portrait element in the portrait picture is worn with the clothing element in the target clothing picture, firstly, the portrait element in the portrait picture and the target clothing picture are needed.
  • the clothing elements are extracted to remove the background of the picture from other irrelevant factors.
  • Digital image processing techniques including image transformation, image enhancement techniques, etc., are required for the extraction of portrait elements and clothing elements in the picture.
  • the conversion of analog image to digital image is realized by image transformation, which facilitates subsequent processing.
  • the methods of image transformation mainly include Fourier transform, Walsh-Hadamard transform and discrete Kafner-Levy transform.
  • Fourier transform is a Fourier transform, including Fourier transform and Fourier inverse transform, Fourier transform is used to transform from time domain to frequency domain, and Fourier inverse transform is used to convert from frequency domain to time domain.
  • the two-dimensional Fourier transform is used to transform from the spatial domain to the transform domain.
  • the transform domain image is the two-dimensional spectrum of the original spatial domain image.
  • the DC term is proportional to the average of the original image brightness, and the high-frequency term is characterized. The intensity and direction of the edge changes in the image.
  • the Walsh-Adamama transform and the discrete Kafner-Levy transform can also realize the conversion function from analog image to digital image.
  • the implementation method of Walsh-Adamama transform is relative to Fu.
  • the inner-leaf transformation is simpler, the computer has higher computing speed and the storage capacity is small;
  • the discrete Kaffner-Levy transform is a transformation based on the statistical properties of the image, also known as the Hotlin transform or the eigenvector transform.
  • the image is transformed from time domain to frequency domain by Fourier transform, Walsh-Hadamard transform and discrete Kafner-Levy transform, there are different frequency components in the image.
  • the image transformation is gentle at low frequency components in the frequency domain.
  • the same block of the same color in the original image is the same or similar and the color block without color mutation is transformed into the frequency domain and then is the low frequency component; the abrupt part or the edge part of the original image is in the frequency domain.
  • the high-frequency components for example, adjacent color patches having large color differences in the original image are converted to the frequency domain and their boundaries correspond to high-frequency components.
  • the image enhancement technology is used to strengthen the high-frequency component of the image, so that the outline in the image is clear, thereby facilitating extraction.
  • the image transformed into the frequency domain is filtered by a high-pass filter, the low-frequency component is suppressed, the high-frequency component is passed, and the high-frequency component is enhanced, thereby enhancing the edge, making the outline of the portrait element and the clothing element clear and easy to extract.
  • the part of the portrait element and the clothing element are distinguished by the established portrait element and the clothing element feature recognition strategy.
  • the above clothing corresponds to the listed torso
  • the distance between the collar and the head is set to the length of the neck and the length of the collar
  • the clothing elements are covered according to the corresponding principles.
  • a composite picture is formed to the corresponding part of the portrait element.
  • the above image processing method may be adopted when referring to the extraction of a portrait element or a clothing element in a picture.
  • the synthesized picture On the display interface of the smart terminal, the synthesized picture is displayed, and the composite picture should include one or more of a front view, a side view, and a rear view, wherein the front view must be included.
  • the user can directly perform virtual fitting on the smart terminal, and the effect of real fitting can be achieved without physical wear, and the matching manner can be selected according to the diversified needs of the user, which is convenient and quick.
  • the step of acquiring a portrait image including a portrait element includes:
  • the camera image can be taken by the camera to obtain a portrait image including a portrait element.
  • the camera in the smart terminal can be directly used to take a photo by acquiring the camera right in the smart terminal, or the camera module can be established in the program design, and the portrait image can be captured by calling the camera of the smart terminal. , thereby realizing the function of taking a picture to obtain a portrait picture including a portrait element.
  • the step of acquiring a portrait image including a portrait element includes:
  • S101 Obtain access rights of an application or a gallery in the smart terminal.
  • S102 Read a portrait image in an application or a gallery in the smart terminal.
  • the step of acquiring a picture of a clothing including a clothing element includes:
  • the camera picture can be taken by the camera to obtain the picture of the clothing including the clothing element, and the camera in the smart terminal can be directly used to take a picture by acquiring the camera right in the smart terminal, or the camera module can be established in the program design, and the picture of the clothing can be captured by calling the camera of the smart terminal. Thereby, a function of taking a picture of a clothing including a clothing element is achieved.
  • the step of acquiring a clothing picture including a clothing element includes:
  • S201 Obtain access rights of an application or a gallery in the smart terminal.
  • S202 Read an image of the clothing in the application or the gallery in the smart terminal.
  • the user can select the clothing picture selected by the user in the application or the gallery.
  • the picture selected by the user does not necessarily fully display the characteristics of the clothing.
  • the picture can be obtained by taking pictures, but the user is shopping.
  • the photographs taken are generally more casual.
  • the characteristics of the clothes cannot be displayed.
  • the quality of the pictures of the clothes is low, including the pixels and resolution of the pictures themselves, and the display of the characteristics of the clothes is not enough, so the optimization module can be provided. Users can decide whether to optimize the picture.
  • the basis for optimizing the clothing picture is to create a gallery containing pictures of clothing brands of various clothing brands, or to link to the official website of each clothing brand by means of hyperlinks, and the program can read and copy the clothing of each clothing brand.
  • the user needs to input the brand name of the selected target clothing, perform a library comparison according to the brand name input by the user and the target clothing picture, and obtain a picture of the clothing containing the same item as the user target clothing. Since the image of the target clothing acquired at this time is a picture for displaying the clothing for each clothing brand, the display of various aspects of the clothing is in most cases higher than the picture of the clothing photographed by the user himself.
  • the target clothing image selected by the user comes from the official website of each brand, so the user can choose not to optimize, and if the user selects optimization at this time, the image quality will not be affected.
  • the acquired picture is a picture of the clothing containing the same item of clothing as the target clothing of the clothing brand.
  • the gallery does not have the same brand of clothing as the user's target clothing.
  • the picture of the object or the picture of the clothing of the same item, the user's target clothing picture will be directly used. Therefore, the clothing picture finally obtained through the application or the library in the smart terminal may be a directly acquired clothing picture or an optimized clothing picture.
  • the function is called to read the above-obtained clothing picture and copy it.
  • the step of generating a composite picture specifically includes:
  • the recognition process includes extracting features, building an index build, and querying. By color sampling the face and the hand, from the hand and the face to the middle color, through the double recognition of the color and the edge mutation, the recognition of the portrait element in the portrait image is achieved. The function of the portion not covered by the clothing is further divided into pictures, and the uncovered portion of the portrait element in the portrait picture is extracted.
  • the standard body type refers to a different standard body type. Such as different height, body proportion, etc.
  • the garment covering portion within the portrait element is reconstructed by the extension of the line and the filling of the color by the standard body type in the uncovered portion of the clothing and the standard body type in the database.
  • the portrait element After reconstructing the clothing covering part in the portrait element, the portrait element is relatively complete. At this time, the clothing element needs to be extracted, and the color and edge image recognition technology is also used in the extraction, and the image is segmented to obtain the clothing element. After the laundry element is extracted, it is necessary to cover the clothing element on the clothing covering portion.
  • Covering the clothing element with the clothing element in order to achieve the effect of the virtual fitting, it is necessary to perform big data analysis on the effect of the real fitting.
  • the effect of dressing including elastic, wide and narrow pants.
  • the figure body includes height, fat and thin.
  • the clothing element is covered on the clothing covering portion according to the corresponding principle, and the effect of the detail portion is adjusted according to the big data analysis result and the figure body and the target clothing size, such as the personal body and the target clothing size and the type of the body. Effect, etc. Thereby, a composite picture in which the target clothing in the target clothing picture is worn by the portrait element in the portrait picture is obtained.
  • the steps of the fitting method further include:
  • S401 Analyze the matching degree between the portrait element and the clothing element in the composite picture to form a first matching degree parameter.
  • the user After obtaining the composite image, the user has a subjective evaluation of the dressing effect. In many cases, the user may need to objectively evaluate the dressing effect of others. Therefore, the method provides an objective evaluation of the user's dressing effect. An objective evaluation is given by analyzing the degree of matching between the portrait element and the clothing element in the composite picture.
  • the matching degree of each skin color and the color of each clothing is established, such as the matching degree of each body type and each clothing type, type, size, etc., taking the dark yellow skin color as an example, the matching degree to white is dark purple High matching.
  • the composite picture is parsed, and then combined with the analysis strategy, the matching degree evaluation of the portrait element and the clothing element in the composite picture is given, and the matching degree is the first matching degree parameter.
  • S402 Push a clothing picture having a second matching degree parameter, where the second matching degree parameter is greater than the first matching degree parameter, based on the first matching degree parameter.
  • the user After forming the first matching degree parameter, the user is pushed to the clothing picture whose matching degree is higher than the first matching degree parameter in combination with the skin color of the user, the body type, and the kind of the clothing element in the target clothing picture.
  • a smart terminal-based fitting system specifically includes the following components:
  • the portrait image acquisition module in order to realize the smart virtual fitting, needs to obtain a portrait image including the portrait elements necessary for trying on the target clothing, the portrait element refers to the main body structure of the human body: the head, the limbs, the trunk, the foot, etc., wherein The upper limb and the upper torso constitute the upper body, the lower limb and the lower torso constitute the lower body, and the portrait element must correspond to the target clothing.
  • the portrait picture may be a half-length photograph including only the head and the upper body; the target clothing is a dress, In a skirt or trousers, the portrait picture must include the body elements of the head, upper body and lower body; when the target clothing is a shoe, the portrait picture must contain the body element of the foot.
  • the target clothing is a shoe. It is also best to include the head, upper body and lower body. Remind the user to upload the corresponding picture when the user picture does not meet the above requirements.
  • the clothing picture acquisition module after obtaining the portrait picture, needs to obtain the target clothing picture including the clothing element, and the quality of the obtained target clothing picture will directly affect the effect of the virtual fitting, and the quality of the target clothing picture here
  • the quantity includes not only the resolution and pixel size of the picture, but also the representation of the picture on the characteristics of the clothing. Therefore, the target clothing picture must contain the complete clothing element, and clearly and completely reveal the characteristics of the clothing. Therefore, the target clothing picture is not only one piece, and the characteristics of the clothing must be exhibited from various directions, including at least the front view and the rear view, which are achieved through the plan view.
  • the three-dimensional effect ensures a better try-on effect.
  • Synthesizing module similarly, when synthesizing the portrait picture and the target clothing picture, so that the portrait element in the portrait picture is wearing the clothing element in the target clothing picture, firstly, the portrait element in the portrait picture is needed And the clothing elements in the target clothing picture are extracted to remove the background of the picture and other irrelevant factors.
  • Digital image processing techniques including image transformation, image enhancement techniques, etc., are required for the extraction of portrait elements and clothing elements in the picture.
  • image transformation which facilitates subsequent processing.
  • the methods of image transformation mainly include Fourier transform, Walsh-Hadamard transform and discrete Kafner-Levy transform.
  • the image is transformed from the time domain to the frequency domain by the Fourier transform, the Walsh-Hadamard transform, and the discrete Kafner-Levy transform, and then transformed into the frequency domain by a high-pass filter.
  • the image is filtered, the low-frequency component is suppressed, the high-frequency component is passed, and the high-frequency component is enhanced, thereby strengthening the edge, so that the outline of the portrait element and the clothing element is clear and easy to extract.
  • it is necessary to segment the portrait image containing the portrait element and the clothing image containing the clothing element, and extract the portrait element and the clothing element.
  • image segmentation can be realized, thereby extracting portrait elements and clothing elements. the goal of.
  • the above clothing corresponds to the listed torso, and the distance between the collar and the head is set to the length of the neck and the length of the collar, and the clothing elements are covered according to the corresponding principles.
  • a composite picture is formed to the corresponding part of the portrait element.
  • the display module displays the synthesized picture on the display interface of the smart terminal, and the composite picture should include one or more of a front view, a side view, and a rear view, wherein the front view must be included.
  • the portrait image acquisition module specifically includes:
  • the camera establishing unit can obtain a portrait picture including a portrait element by photographing the camera, and can directly take a picture by using a camera in the smart terminal by acquiring the camera right in the smart terminal, or can be programmed.
  • the camera module is established, and the portrait image is captured by calling the camera of the smart terminal, thereby realizing the function of taking a portrait image including the portrait element.
  • the photo reading unit calls a function to read a portrait picture taken by the camera.
  • the portrait image acquisition module specifically includes:
  • the read permission acquisition unit because it needs to retrieve the portrait image from the application or the library in the smart terminal, involves the deep content of the application or the gallery, so it is necessary to first make a request to the application to obtain the permission to read the image.
  • the photo reading unit calls a function to read and copy the portrait image in the application or the gallery in the smart terminal.
  • the clothing picture obtaining module specifically includes:
  • the camera establishing unit can obtain a picture of the clothing including the clothing element by photographing the camera, and can directly take a picture by using the camera in the smart terminal by acquiring the camera right in the smart terminal, or can establish a camera module in the program design, by calling the smart terminal.
  • the camera captures a picture of the laundry, thereby realizing the function of taking a picture of the clothing including the clothing element.
  • the photo reading unit calls a function to read the image of the clothing taken by the camera and copy it.
  • the clothing picture obtaining module specifically includes:
  • the read permission acquisition unit because it needs to retrieve the clothing picture from the application or the library in the smart terminal, involves the deep content of the application or the gallery, so it is necessary to first make a request to the application to obtain the permission to read the picture.
  • the photo reading unit can directly read the clothing picture selected by the user in the application or the gallery after obtaining the access right of the application or the gallery in the smart terminal.
  • the picture selected by the user does not necessarily fully display the characteristics of the clothing.
  • the picture can be obtained by taking pictures, but the user is shopping.
  • the photographs taken are generally more casual.
  • the characteristics of the clothes cannot be displayed.
  • the quality of the pictures of the clothes is low, including the pixels and resolution of the pictures themselves, and the display of the characteristics of the clothes is not enough, so the optimization module can be provided. Users can decide whether to optimize the picture.
  • the basis for optimizing the clothing picture is to create a gallery containing pictures of clothing brands of various clothing brands, or to link to the official website of each clothing brand by means of hyperlinks, and the program can read and copy the clothing of each clothing brand.
  • the user needs to input the brand name of the selected target clothing, perform a library comparison according to the brand name input by the user and the target clothing picture, and obtain a picture of the clothing containing the same item as the user target clothing. Since the image of the target clothing acquired at this time is a picture for displaying the clothing for each clothing brand, the display of various aspects of the clothing is in most cases higher than the picture of the clothing photographed by the user himself.
  • the target clothing image selected by the user comes from the official website of each brand, so the user can choose not to optimize, and if the user chooses optimization at this time, the picture quality will not be lowered. Low impact.
  • the acquired picture is a picture of the clothing containing the same item of clothing as the target clothing of the clothing brand.
  • the target clothing picture provided by the user is directly used. Therefore, the clothing picture finally obtained through the application or the library in the smart terminal may be a directly acquired clothing picture or an optimized clothing picture. The function is called to read the above-obtained clothing picture and copy it.
  • the synthesizing module specifically includes:
  • the portrait element capturing unit needs to ensure that the portrait element in the user portrait image is restored as much as possible, and the portrait image has a part of the portrait element covered by the clothing. Therefore, the portrait image needs to be captured first.
  • the recognition process includes extracting features, building an index build, and querying. By color sampling the face and the hand, from the hand and the face to the middle color, through the double recognition of the color and the edge mutation, the recognition of the portrait element in the portrait image is achieved.
  • the function of the portion not covered by the clothing is further divided into pictures, and the uncovered portion of the portrait element in the portrait picture is extracted.
  • Portrait contour reconstruction unit After capturing the uncovered portion of the portrait element in the portrait image, in order to restore the portrait element, it is necessary to reconstruct the clothing covering portion within the portrait element.
  • the standard body type refers to different standard body types, such as different height and body proportion, etc.
  • Covering unit after reconstructing the clothing covering part in the portrait element, the portrait element is relatively complete, and the clothing element needs to be extracted at this time, and the image recognition technology of color and edge is also adopted in the extraction, and the image is segmented to obtain the clothing. element.
  • the laundry element After the laundry element is extracted, it is necessary to cover the clothing element on the clothing covering portion.
  • big data analysis is needed for the effect of real fitting. Create a large database of real fitting images, and analyze the real fitting effect big data, such as different materials, different patterns and different lengths of different pants for different types of people, including elastic, pants The width of the mouth is equal.
  • the figure body includes height, fat and thin.
  • the target clothing Covering the clothing element with the clothing element in accordance with the corresponding principle, and analyzing the result according to big data and the figure of the character, the target clothing
  • the effect of the size adjustment details such as the body effect and the size of the target clothing and the fit effect caused by the type.
  • a composite picture in which the target clothing in the target clothing picture is worn by the portrait element in the portrait picture is obtained.
  • the fitting system further comprises:
  • the matching degree analysis module obtains a subjective evaluation of the dressing effect after obtaining the synthesized picture. In many cases, the user may need to objectively evaluate the dressing effect of others. Therefore, the method provides an objective evaluation of the user's dressing effect.
  • An objective evaluation is given by analyzing the degree of matching between the portrait element and the clothing element in the composite picture. First of all, it is necessary to establish an analysis strategy, such as the matching degree of each skin color and the color of each clothing, the matching degree of each body type and each clothing type, type, size, etc., taking the dark yellow skin color as an example, the matching degree to white is dark purple High matching.
  • the composite picture is parsed, and then combined with the analysis strategy, the matching degree evaluation of the portrait element and the clothing element in the composite picture is given, and the matching degree is the first matching degree parameter.
  • the clothing picture pushing module after forming the first matching degree parameter, pushes the user with the clothing picture whose matching degree is higher than the first matching degree parameter in combination with the skin color of the user, the body shape, and the type of the clothing element in the target clothing picture.

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Hardware Design (AREA)
  • Evolutionary Computation (AREA)
  • Geometry (AREA)
  • General Engineering & Computer Science (AREA)
  • Image Processing (AREA)
  • Processing Or Creating Images (AREA)
  • Image Analysis (AREA)

Abstract

一种基于智能终端的试衣方法及试衣系统，试衣方法包括以下步骤：获取包括人像元素的人像图片；获取包括衣物元素的目标衣物图片；将所述人像图片与所述目标衣物图片合成形成一合成图片，使所述人像图片中的人像元素穿戴有所述目标衣物图片中的衣物元素；显示所述合成图片。采用上述方法后，方便智能实现虚拟试衣，无需实体穿戴试衣即可达到实体试衣效果。

Description

一种基于智能终端的试衣方法及试衣系统 技术领域
本发明涉及智能终端技术领域，尤其涉及一种基于智能终端的试衣方法及试衣系统。
背景技术
人们在逛街时，常常看到喜欢的衣服但不方便试穿，尤其是在浏览购物网站时，无法试穿同时又担心衣服不合适，导致购买衣物耗时长、满意度低，网站购物还存在大小不合适、退货率高的可能性。随着信息技术的发展，已经存在一些虚拟试衣的系统，通过用户提供肖像照片与待试衣服的照片进行叠加，形成虚拟试衣的图片，供用户参考。但其衣物照片来源单一，衣物图片的质量不高，无法再现衣物的特征，同时，目前的试衣效果图仅为正面，不够完整。本发明通过根据用户提供的衣物照片智能替换由相应品牌官网提供的同款衣物照片，并给出正面、背面、侧面的试衣效果图，更好地展现试衣效果，尽可能模拟真实试衣，并且可以根据用户的肤色、身材等信息与试穿衣物的匹配度给出专业评价，智能推荐匹配度更高衣物照片。
发明内容
为了克服上述技术缺陷，本发明的目的在于提供一种基于智能终端的试衣方法及试衣系统，可实现多种情况下的虚拟试衣，并提供最佳的虚拟试衣方案。
本发明公开了一种基于智能终端的试衣方法，包括以下步骤：获取包括人像元素的人像图片；获取包括衣物元素的目标衣物图片；将所述人像图片与所述目标衣物图片合成形成一合成图片，使所述人像图片中的人像元素穿戴有所述目标衣物图片中的衣物元素；显示所述合成图片。
优选地，获取包括人像元素的人像图片的方法包括：通过照相机拍照获取人像图片，或获取所述智能终端内应用程序或图库内的人像图片；当获取人像图片的方法为通过照相机拍照获取时，其步骤包括：获取智能终端内的照相机权限或于智能终端内建立照相机模块；读取照相机拍取的人像图片；当获取人像图片的方法为获取所述智能终端内应用程序或图库内的人像图片时，其步骤包括：获取所述智能终端内应用程序或图库的访问权限；读取所述智能终端内应用程序或图库内的人像图片。
优选地，获取包括衣物元素的目标衣物图片的方法包括：通过照相机拍照获取衣物图片、或获取所述智能终端内应用程序或图库内的衣物图片；当获取衣物图片的方法为通过照相机拍照获取时，其步骤包括：获取智能终端内的照相机权限或于智能终端内建立照相机模块；读取照相机拍取的衣物图片；当获取衣物图片的方法为获取所述智能终端内应用程序或图库内的衣物图片时，其步骤包括：获取所述智能终端内应用程序或图库的访问权限；读取所述智能终端内应用程序或图库内的衣物图片。
优选地，将所述人像图片与所述目标衣物图片合成形成一合成图片，使所述人像图片中的人像元素穿戴有所述目标衣物图片中的衣物元素的步骤包括：捕捉所述人像图片中人像元素内的未被衣物覆盖部分；根据所述未被衣物覆盖部分重构所述人像元素内的衣物覆盖部分的轮廓；将所述衣物元素覆盖于所述衣物覆盖部分上。
优选地，在显示所述合成图片的步骤后，所述方法还包括：分析所述合成图片中人像元素与衣物元素的匹配度，形成一第一匹配度参数；基于所述第一匹配度参数，推送一具有第二匹配度参数的衣物图片，其中所述第二匹配度参数大于所述第一匹配度参数。
本发明还公开了一种基于智能终端的试衣系统，包括：人像图片获取模块，获取包含人像元素的人像图片；衣物图片获取模块，获取包含衣物元素的目标衣物图片；合成模块，将所述人像图片与所述目标衣物图片合成形成一合成图片，使所述人像图片中的人像元素穿戴有所述目标衣物图片中的衣物元素；显示模块，显示所述合成图片。
优选地，所述人像图片获取模块包括：照相机建立单元，调用智能终端摄像头捕获并固定图片；照片读取单元，读取照相机内的人像图片；或所述人像图片获取模块包括：读取权限获取单元，向所述智能终端发送一权限请求，以获取所述智能终端内应用程序、图库读取权限；照片读取单元，读取应用程序或图库内的人像图片。
优选地，所述衣物图片获取模块包括：照相机建立单元，调用智能终端摄像头捕获并固定图片；照片读取单元，读取照相机内的衣物图片；或所述衣物图片获取模块包括：读取权限获取单元，向所述智能终端发送一权限请求，以获取所述智能终端内应用程序、图库读取权限；照片读取单元，读取应用程序或图库内的衣物图片。
优选地，所述合成模块包括：人像元素捕捉单元，捕捉所述人像图片中人像元素内的未被衣物覆盖部分；人像轮廓重构单元，根据所述未被衣物覆盖部分重构所述人像元素内的衣物覆盖部分的轮廓；覆盖单元，将所述衣物元素覆盖于所述衣物覆盖部分上。
优选地，在显示模块后，所述系统还包括：匹配度分析模块，分析所述合成图片中人像元素与衣物元素的匹配度，形成一第一匹配度参数；衣物图片推送模块，基于所述 第一匹配度参数，推送一具有第二匹配度参数的衣物图片，其中所述第二匹配度参数大于所述第一匹配度参数。
采用了上述技术方案后，与现有技术相比，具有以下有益效果：
1.通过选取最优的目标衣物图片，达到最好的虚拟试衣效果；
2.完整地展现多个视角的试衣效果；
3.使用户对试衣效果有充分且客观的了解；
4.帮助用户选取到最符合用户需求的衣物。
附图说明
图1为符合本发明一优选实施例中试衣方法的流程示意图；
图2为符合本发明一优选实施例中，试衣方法的获取人像图片的方法的流程示意图；
图3为符合本发明另一优选实施例中，试衣方法的获取人像图片的方法的流程示意图；
图4为符合本发明一优选实施例中，试衣方法的获取目标衣物的方法的流程示意图；
图5为符合本发明另一优选实施例中，试衣方法的获取目标衣物的另一方法的流程示意图；
图6为符合本发明一优选实施例中，试衣方法的生成合成图片的方法的流程示意图；
图7为符合本发明另一优选实施例中试衣方法的流程示意图；
图8为符合本发明一优选实施例中试衣系统的系统结构示意图。
图9为符合本发明一优选实施例中，试衣系统的人像图片获取模块的结构示意图。
图10为符合本发明另一优选实施例中，试衣系统的人像图片获取模块的结构示意图。
图11为符合本发明一优选实施例中，试衣系统的衣物图片获取模块的结构示意图。
图12为符合本发明另一优选实施例中，试衣系统的衣物图片获取模块的结构示意图。
图13为符合本发明一优选实施例中，试衣系统的合成模块的结构示意图。
图14为符合本发明另一优选实施例中试衣系统的结构示意图。
具体实施方式
这里将详细地对示例性实施例进行说明，其示例表示在附图中。下面的描述涉及附图时，除非另有表示，不同附图中的相同数字表示相同或相似的要素。以下示例性实施例中所描述的实施方式并不代表与本公开相一致的所有实施方式。相反，它们仅是与如所附权利要求书中所详述的、本公开的一些方面相一致的装置和方法的例子。
在本公开使用的术语是仅仅出于描述特定实施例的目的，而非旨在限制本公开。在本公开和所附权利要求书中所使用的单数形式的“一种”、“所述”和“该”也旨在包括多数形式，除非上下文清楚地表示其他含义。还应当理解，本文中使用的术语“和/或”是指并包含一个或多个相关联的列出项目的任何或所有可能组合。
在本发明的描述中，除非另有规定和限定，对于本领域的普通技术人员而言，可以根据具体情况理解术语的具体含义。
在后续的描述中，使用用于表示元件的诸如“模块”、“部件”或“单元”的后缀仅为了有利于本发明的说明，其本身并没有特定的意义，仅表示系统的一部分。因此，“模块”与“部件”可以混合地使用。
智能终端可以以各种形式来实施。例如，本发明中描述的终端可以包括诸如移动电话、智能电话、笔记本电脑、PDA(个人数字助理)、PAD(平板电脑)、PMP(便携式多媒体播放器)、导航装置等等的移动终端以及诸如数字TV、台式计算机等等的固定终端。
参阅图1，为本发明一优选实施例中基于智能终端的试衣方法的流程示意图。该实施例中，试衣方法具体包括以下步骤：
S100：获取包括人像元素的人像图片
为了实现智能虚拟试衣，需要获取包含试穿目标衣物所必需的人像元素的人像图片，人像元素是指人体主要身体结构：头部、四肢、躯干、脚部等，其中，上肢与上身躯干构成上半身，下肢与下身躯干构成下半身，人像元素必须与目标衣物相对应，如：目标衣物为上衣时，人像图片可为仅包含头部以及上半身的半身照；目标衣物为连衣裙、半身裙或裤子时，人像图片则必须包含头部、上半身以及下半身在内的身体元素；目标衣物为鞋子时，人像图片必须包含脚部这一身体元素，为考虑整体视觉效果，目标衣物为鞋子时也最好包含头部、上半身以及下半身。当用户图片不满足上述要求时提醒用户上传相应的图片。
S200：获取包括衣物元素的目标衣物图片
获取人像图片后，需要获取包括衣物元素的目标衣物图片，获取的目标衣物图片的质量高低将直接影响到虚拟试衣的效果，这里的目标衣物图片的质量不仅包括图片的分辨率与像素大小，还包括图片对衣物特征的展现情况。因此，目标衣物图片必须包含完整的衣物元素，并且清楚完整地展现衣物的特征，因此，目标衣物图片不只一张，必须从各个方位展现衣物的特征，至少包括正面图与背面图，通过平面图达到三维立体的效果，从而保证更好的试穿效果。
S300：将所述人像图片与所述目标衣物图片合成形成一合成图片，使所述人像图片中的人像元素穿戴有所述目标衣物图片中的衣物元素
在对人像图片与所述目标衣物图片进行合成，使所述人像图片中的人像元素穿戴有所述目标衣物图片中的衣物元素时，首先需要对人像图片中的人像元素以及目标衣物图片中的衣物元素进行提取，去除图片的背景与其它不相干因素。
对图片中人像元素以及衣物元素提取需要采用数字图像处理技术，包括图像变换、图像增强技术等。
首先，通过图像变换实现模拟图像到数字图像的转换，方便后续的处理，图像变换的方法主要有傅里叶变换、沃尔什-阿达玛变换以及离散卡夫纳－勒维变换。傅里叶变换即Fourier变换，包括傅里叶正变换与傅里叶逆变换，傅里叶正变换实现从时域到频域的转换，傅里叶逆变换实现从频域到时域的转换，此处采用二维傅里叶正变换，从空间域变换到转换域，转换域图像是原空间域图像的二维频谱，其直流项与原图像亮度的平均值成比例，高频项表征图像中边缘变化的强度和方向。除傅里叶变换外，沃尔什-阿达玛变换与离散卡夫纳－勒维变换也可以实现模拟图像到数字图像的转换功能，其中，沃尔什-阿达玛变换的实现方法相对与傅里叶变换更简单，计算机的运算速度更高，存储容量小；离散卡夫纳－勒维变换是以图像的统计特性为基础的变换，又称霍特林变换或本征向量变换。
通过傅里叶变换、沃尔什-阿达玛变换以及离散卡夫纳－勒维变换中的任一种变换方法实现图像从时域变换到频域后，在图像中存在不同的频率分量，原图像变换平缓处为频域中的低频分量，如，原图像中同一块颜色相同或相近且没有颜色突变的色块变换到频域后为低频分量；原图像突变部分或边缘部分为频域中的高频分量，如，原图像中存在颜色差别较大的相邻色块，则转换至频域后其边界对应高频分量。因此，在提取相应的人像元素或衣物元素时，采用图像增强技术，强化图像高频分量，使图像中轮廓清晰，从而方便提取。此处，通过高通滤波器对变换到频域的图像进行滤波，抑制低频分量，使高频分量通过，实现高频分量的增强，从而强化边缘，使人像元素与衣物元素轮廓清晰，便于提取。
强化人像元素与衣物元素轮廓后，需要对含人像元素的人像图片与含衣物元素的衣物图片进行分割，提取人像元素与衣物元素。根据图像中像素的亮度及颜色进行分割，包括灰度阀值分割法、区域生长分割法、边缘分割法以及基于直方图递归分割法，均可实现图像的分割，从而达到提取人像元素与衣物元素的目的。
建立人像元素以及衣物元素特征识别策略，在此过程中用到图片识别技术，包括提取特征、建立索引build以及查询三个步骤，通过对人像的头部、四肢、躯干、脚部的特征进行判断，如：长度、宽度以及各部分比例，实现对人像的头部、四肢、躯干、脚部各部分的区分；通过对衣物的区别特征的设置，实现对上衣、连衣裙、裤子等不同种类衣物的区分。
提取人像元素与衣物元素后，通过所建立的人像元素以及衣物元素特征识别策略对与人像元素与衣物元素各部分进行识别区分。预先设置的各部分人像元素与衣物元素对应原则，如上衣与上市躯干对应，并设置上衣衣领到头部之间的距离为脖子的长度与衣领的长度差，根据对应原则将衣物元素覆盖至人像元素的相应部分，形成合成图片。
同样的，在其他实施实施例中，涉及到图片中人像元素或衣物元素的提取时均可采取上述图像处理方法。
S400：显示所述合成图片
在智能终端的显示界面上，显示已经合成图片，合成图片应包括正面图、侧面图以及背面图中的一种或多种，其中，必须包含正面图。
通过上述方法，用户可直接在智能终端上进行虚拟试衣，无需实体穿戴即可达到真实试衣的效果，并且可以根据用户多元化的需求自行选择搭配方式，方便快捷。
参阅图2，在一优选实施例中，获取包括人像元素的人像图片的步骤，具体包括：
S101：获取智能终端内的照相机权限或于智能终端内建立照相机模块
通过照相机拍照可以获取包括人像元素的人像图片，可以通过获取智能终端内的照相机权限直接利用智能终端内的照相机进行拍照，也可以在程序设计中建立照相机模块，通过调用智能终端的摄像头捕获人像图片，从而实现拍照获取包括人像元素的人像图片的功能。
S102：读取照相机拍取的人像图片
调用函数，对照相机拍取的人像图片进行读的操作，并进行复制。
参阅图3，在另一优选实施例中，获取包括人像元素的人像图片的步骤，具体包括：
S101：获取所述智能终端内应用程序或图库的访问权限
由于需要从智能终端内应用程序或图库中调取人像图片，涉及到应用程序或图库的深层内容，因此需先向应用程序发出请求，获取读取其图片的权限。
S102：读取所述智能终端内应用程序或图库内的人像图片
调用函数，对所述智能终端内应用程序或图库内的人像图片进行读的操作，并进行复制。
参阅图4，在一优选实施例中，获取包括衣物元素的衣物图片的步骤，具体包括：
S201：获取智能终端内的照相机权限或于智能终端内建立照相机模块
通过照相机拍照可以获取包括衣物元素的衣物图片，可以通过获取智能终端内的照相机权限直接利用智能终端内的照相机进行拍照，也可以在程序设计中建立照相机模块，通过调用智能终端的摄像头捕获衣物图片，从而实现拍照获取包括衣物元素的衣物图片的功能。
S202：读取照相机拍取的衣物图片
调用函数，对照相机拍取的衣物图片进行读的操作，并进行复制。
参阅图5，在另一优选实施例中，获取包括衣物元素的衣物图片的步骤，具体包括：
S201：获取所述智能终端内应用程序或图库的访问权限
由于需要从智能终端内应用程序或图库中调取衣物图片，涉及到应用程序或图库的深层内容，因此需先向应用程序发出请求，获取读取其图片的权限。
S202：读取所述智能终端内应用程序或图库内的衣物图片
获取所述智能终端内应用程序或图库的访问权限后，可以直接读取应用程序或图库内用户选定的衣物图片。但用户选定的衣物图片并不一定能够完整地展现衣物的特征，很多时候，当用户在逛街的时候不想或不方便试穿衣物时，其可以通过拍照获得衣物图片，但用户在逛街时所拍的照片一般较为随意，绝大多数情况下都不能展现衣物的特征，衣物图片质量较低，既包括图片自身的像素与分辨率，也包括对衣物特征的展现不够，因此可提供优化模块，用户可自行决定是否优化图片。
优化衣物图片的基础在于建立包含各服装品牌衣物图片的图库，或通过超链接的方式链接到各服装品牌官方网站，程序可以对各服装品牌衣物进行读与复制操作。用户需要输入其所选定目标衣物所属品牌名称，根据用户输入的品牌名称以及目标衣物图片进行图库比对，获取包含与用户目标衣物同款衣物的衣物图片。由于此时获取的目标衣物图片为各服装品牌拍摄用来展示衣物的图片，对衣物各方面特征的展现绝大多数情况下都比用户自行拍摄的衣物图片质量高。不排除用户选定的目标衣物图片本身来源于各品牌官网，因此用户可以选择不进行优化，如果此时用户选择优化也不会造成图片质量降低的影响。在优化衣物图片后，所获取的照片即为由各服装品牌专业拍摄包含与用户目标衣物同款衣物的衣物图片。当然，也有可能图库中并没有与用户目标衣物同品牌的衣 物图片或同款衣物的衣物图片，此时会直接使用用户提供的目标衣物图片。因此，最终通过智能终端内应用程序或图库获取的衣物图片既可能是直接获取的衣物图片，也可能是优化后的衣物图片。
调用函数，对上述最终获取的衣物图片进行读的操作，并进行复制。
参阅图6，在一优选实施例中，生成合成图片的步骤具体包括：
S301：捕捉所述人像图片中人像元素内的未被衣物覆盖部分
为了真实地展现用户的试衣效果，需要保证用户人像图片中人像元素被尽可能还原，而人像图片中存在一部分被衣物覆盖的人像元素，因此，需要首先捕捉所述人像图片中人像元素内的未被衣物覆盖部分。
为了捕捉所述人像图片中人像元素内的未被衣物覆盖部分，需要利用图像识别技术，对于皮肤的识别采取色彩与边缘识别的策略。识别过程包括提取特征、建立索引build以及查询，通过对面部、手部进行颜色取样，由手部、面部向中间颜色，通过颜色与边缘突变进行双识别，达到识别所述人像图片中人像元素内的未被衣物覆盖部分的功能，再进行图片分割，提取所述人像图片中人像元素内的未被衣物覆盖部分。
S302：根据所述未被衣物覆盖部分重构所述人像元素内的衣物覆盖部分的轮廓
捕捉所述人像图片中人像元素内的未被衣物覆盖部分后，为了对人像元素进行还原，需要重构所述人像元素内的衣物覆盖部分。
对于根据所述未被衣物覆盖部分重构所述人像元素内的衣物覆盖部分的轮廓，首先需要通过多种体型人群的采样建立多种标准体型数据库，此时的标准体型是指不同标准的体型，如不同的身高、身材比例等；其次需要对人像图片中人像元素进行分析，获取其体型特征，在多种标准体型数据库中匹配与之最接近的标准体型；结合人像图片中人像元素内的未被衣物覆盖部分与数据库中的标准体型对未被衣物覆盖部分进行线条的延伸与颜色的填充，重构所述人像元素内的衣物覆盖部分。
S303：将所述衣物元素覆盖于所述衣物覆盖部分上
重构所述人像元素内的衣物覆盖部分后，人像元素已经比较完整，此时需要对衣物元素进行提取，提取时同样采用色彩与边缘的图像识别技术，并进行图像分割，获取衣物元素。提取衣物元素后，需要将衣物元素覆盖于所述衣物覆盖部分上。
将衣物元素覆盖于所述衣物覆盖部分，为了达到虚拟试衣的效果，需要对真实试衣的效果进行大数据分析。建立大量真实试衣图片的数据库，对真实试衣效果进行大数据分析，如试穿裤子时不同材质、不同版型以及不同长短的不同裤子对于不同体型的人群 的穿衣效果，包括松紧、裤口的宽窄等。其次，需要获取人物身材、目标衣物尺寸，人物身材包括身高、胖瘦等。结合对应原则将所述衣物元素覆盖于所述衣物覆盖部分上，并根据大数据分析结果以及人物身材、目标衣物尺寸调整细节部分的效果，如由人物身材与目标衣物尺寸以及版型导致的贴身效果等。由此，获得由人像图片中的人像元素穿戴目标衣物图片中的目标衣物的合成图片。
参阅图7，在另一优选实施例中，试衣方法的步骤还包括：
S401：分析所述合成图片中人像元素与衣物元素的匹配度，形成一第一匹配度参数
获取合成图片后，用户自身对穿衣效果存在主观的评价，很多时候用户可能需要对他人对穿衣效果的客观评价，因此，本方法提供对用户穿衣效果的客观评价。通过分析所述合成图片中人像元素与衣物元素的匹配度，给出客观评价。
首先需要建立分析策略，如各肤色与各衣物颜色的匹配度、各体型与各衣物种类、版型、尺寸的匹配度等，以暗黄肤色为例，其对白色的匹配度比对暗紫色的匹配度高。其次，对合成图片进行解析，再结合分析策略，给出所述合成图片中人像元素与衣物元素的匹配度评价，此匹配度为第一匹配度参数。
S402：基于所述第一匹配度参数，推送一具有第二匹配度参数的衣物图片，其中所述第二匹配度参数大于所述第一匹配度参数
在形成第一匹配度参数后，结合用户的肤色、体型以及目标衣物图片中衣物元素的种类给用户推送匹配度高于第一匹配度参数的衣物图片。
参阅图8，为符合本发明一优选实施例中基于智能终端的试衣系统，其具体包括以下部件：
人像图片获取模块，为了实现智能虚拟试衣，需要获取包含试穿目标衣物所必需的人像元素的人像图片，人像元素是指人体主要身体结构：头部、四肢、躯干、脚部等，其中，上肢与上身躯干构成上半身，下肢与下身躯干构成下半身，人像元素必须与目标衣物相对应，如：目标衣物为上衣时，人像图片可为仅包含头部以及上半身的半身照；目标衣物为连衣裙、半身裙或裤子时，人像图片则必须包含头部、上半身以及下半身在内的身体元素；目标衣物为鞋子时，人像图片必须包含脚部这一身体元素，为考虑整体视觉效果，目标衣物为鞋子时也最好包含头部、上半身以及下半身。当用户图片不满足上述要求时提醒用户上传相应的图片。
衣物图片获取模块，获取人像图片后，需要获取包括衣物元素的目标衣物图片，获取的目标衣物图片的质量高低将直接影响到虚拟试衣的效果，这里的目标衣物图片的质 量不仅包括图片的分辨率与像素大小，还包括图片对衣物特征的展现情况。因此，目标衣物图片必须包含完整的衣物元素，并且清楚完整地展现衣物的特征，因此，目标衣物图片不只一张，必须从各个方位展现衣物的特征，至少包括正面图与背面图，通过平面图达到三维立体的效果，从而保证更好的试穿效果。
合成模块，同样的，在对人像图片与所述目标衣物图片进行合成，使所述人像图片中的人像元素穿戴有所述目标衣物图片中的衣物元素时，首先需要对人像图片中的人像元素以及目标衣物图片中的衣物元素进行提取，去除图片的背景与其它不相干因素。
对图片中人像元素以及衣物元素提取需要采用数字图像处理技术，包括图像变换、图像增强技术等。首先，通过图像变换实现模拟图像到数字图像的转换，方便后续的处理，图像变换的方法主要有傅里叶变换、沃尔什-阿达玛变换以及离散卡夫纳－勒维变换。通过傅里叶变换、沃尔什-阿达玛变换以及离散卡夫纳－勒维变换中的任一种变换方法实现图像从时域变换到频域后，通过高通滤波器对变换到频域的图像进行滤波，抑制低频分量，使高频分量通过，实现高频分量的增强，从而强化边缘，使人像元素与衣物元素轮廓清晰，便于提取。强化人像元素与衣物元素轮廓后，需要对含人像元素的人像图片与含衣物元素的衣物图片进行分割，提取人像元素与衣物元素。根据图像中像素的亮度及颜色进行分割，包括灰度阀值分割法、区域生长分割法、边缘分割法以及基于直方图递归分割法，均可实现图像的分割，从而达到提取人像元素与衣物元素的目的。利用图片识别技术建立人像元素以及衣物元素特征识别策略，包括提取特征、建立索引build以及查询三个步骤，通过对人像的头部、四肢、躯干、脚部的特征进行判断，如：长度、宽度以及各部分比例，实现对人像的头部、四肢、躯干、脚部各部分的区分；通过对衣物的区别特征的设置，实现对上衣、连衣裙、裤子等不同种类衣物的区分。提取人像元素与衣物元素后，通过所建立的人像元素以及衣物元素特征识别策略对与人像元素与衣物元素各部分进行识别区分。预先设置的各部分人像元素与衣物元素对应原则，如上衣与上市躯干对应，并设置上衣衣领到头部之间的距离为脖子的长度与衣领的长度差，根据对应原则将衣物元素覆盖至人像元素的相应部分，形成合成图片。
显示模块，在智能终端的显示界面上，显示已经合成图片，合成图片应包括正面图、侧面图以及背面图中的一种或多种，其中，必须包含正面图。
参阅图9，在一优选实施例中，人像图片获取模块具体包括：
照相机建立单元，通过照相机拍照可以获取包括人像元素的人像图片，可以通过获取智能终端内的照相机权限直接利用智能终端内的照相机进行拍照，也可以在程序设计 中建立照相机模块，通过调用智能终端的摄像头捕获人像图片，从而实现拍照获取包括人像元素的人像图片的功能。
照片读取单元，调用函数，对照相机拍取的人像图片进行读的操作。
参阅图10，在另一优选实施例中，人像图片获取模块具体包括：
读取权限获取单元，由于需要从智能终端内应用程序或图库中调取人像图片，涉及到应用程序或图库的深层内容，因此需先向应用程序发出请求，获取读取其图片的权限。
照片读取单元，调用函数对所述智能终端内应用程序或图库内的人像图片进行读的操作，并进行复制。
参阅图11，在一优选实施例中，衣物图片获取模块具体包括：
照相机建立单元，通过照相机拍照可以获取包括衣物元素的衣物图片，可以通过获取智能终端内的照相机权限直接利用智能终端内的照相机进行拍照，也可以在程序设计中建立照相机模块，通过调用智能终端的摄像头捕获衣物图片，从而实现拍照获取包括衣物元素的衣物图片的功能。
照片读取单元，调用函数，对照相机拍取的衣物图片进行读的操作，并进行复制。
参阅图12，在另一优选实施例中，衣物图片获取模块具体包括：
读取权限获取单元，由于需要从智能终端内应用程序或图库中调取衣物图片，涉及到应用程序或图库的深层内容，因此需先向应用程序发出请求，获取读取其图片的权限。
照片读取单元，获取所述智能终端内应用程序或图库的访问权限后，可以直接读取应用程序或图库内用户选定的衣物图片。但用户选定的衣物图片并不一定能够完整地展现衣物的特征，很多时候，当用户在逛街的时候不想或不方便试穿衣物时，其可以通过拍照获得衣物图片，但用户在逛街时所拍的照片一般较为随意，绝大多数情况下都不能展现衣物的特征，衣物图片质量较低，既包括图片自身的像素与分辨率，也包括对衣物特征的展现不够，因此可提供优化模块，用户可自行决定是否优化图片。
优化衣物图片的基础在于建立包含各服装品牌衣物图片的图库，或通过超链接的方式链接到各服装品牌官方网站，程序可以对各服装品牌衣物进行读与复制操作。用户需要输入其所选定目标衣物所属品牌名称，根据用户输入的品牌名称以及目标衣物图片进行图库比对，获取包含与用户目标衣物同款衣物的衣物图片。由于此时获取的目标衣物图片为各服装品牌拍摄用来展示衣物的图片，对衣物各方面特征的展现绝大多数情况下都比用户自行拍摄的衣物图片质量高。不排除用户选定的目标衣物图片本身来源于各品牌官网，因此用户可以选择不进行优化，如果此时用户选择优化也不会造成图片质量降 低的影响。在优化衣物图片后，所获取的照片即为由各服装品牌专业拍摄包含与用户目标衣物同款衣物的衣物图片。当然，也有可能图库中并没有与用户目标衣物同品牌的衣物图片或同款衣物的衣物图片，此时会直接使用用户提供的目标衣物图片。因此，最终通过智能终端内应用程序或图库获取的衣物图片既可能是直接获取的衣物图片，也可能是优化后的衣物图片。调用函数，对上述最终获取的衣物图片进行读的操作，并进行复制。
参阅图13，在一优选实施例中，合成模块具体包括：
人像元素捕捉单元，为了真实地展现用户的试衣效果，需要保证用户人像图片中人像元素被尽可能还原，而人像图片中存在一部分被衣物覆盖的人像元素，因此，需要首先捕捉所述人像图片中人像元素内的未被衣物覆盖部分。首先，需要利用图像识别技术，对于皮肤的识别采取色彩与边缘识别的策略。识别过程包括提取特征、建立索引build以及查询，通过对面部、手部进行颜色取样，由手部、面部向中间颜色，通过颜色与边缘突变进行双识别，达到识别所述人像图片中人像元素内的未被衣物覆盖部分的功能，再进行图片分割，提取所述人像图片中人像元素内的未被衣物覆盖部分。
人像轮廓重构单元。捕捉所述人像图片中人像元素内的未被衣物覆盖部分后，为了对人像元素进行还原，需要重构所述人像元素内的衣物覆盖部分。首先需要通过多种体型人群的采样建立多种标准体型数据库，此时的标准体型是指不同标准的体型，如不同的身高、身材比例等；其次需要对人像图片中人像元素进行分析，获取其体型特征，在多种标准体型数据库中匹配与之最接近的标准体型；结合人像图片中人像元素内的未被衣物覆盖部分与数据库中的标准体型对未被衣物覆盖部分进行线条的延伸与颜色的填充，重构所述人像元素内的衣物覆盖部分。
覆盖单元，重构所述人像元素内的衣物覆盖部分后，人像元素已经比较完整，此时需要对衣物元素进行提取，提取时同样采用色彩与边缘的图像识别技术，并进行图像分割，获取衣物元素。提取衣物元素后，需要将衣物元素覆盖于所述衣物覆盖部分上。为了达到虚拟试衣的效果，需要对真实试衣的效果进行大数据分析。建立大量真实试衣图片的数据库，对真实试衣效果进行大数据分析，如试穿裤子时不同材质、不同版型以及不同长短的不同裤子对于不同体型的人群的穿衣效果，包括松紧、裤口的宽窄等。其次，需要获取人物身材、目标衣物尺寸，人物身材包括身高、胖瘦等。结合对应原则将所述衣物元素覆盖于所述衣物覆盖部分上，并根据大数据分析结果以及人物身材、目标衣物 尺寸调整细节部分的效果，如由人物身材与目标衣物尺寸以及版型导致的贴身效果等。由此，获得由人像图片中的人像元素穿戴目标衣物图片中的目标衣物的合成图片。
参阅图14，在另一优选实施例中，试衣系统还包括：
匹配度分析模块，获取合成图片后，用户自身对穿衣效果存在主观的评价，很多时候用户可能需要对他人对穿衣效果的客观评价，因此，本方法提供对用户穿衣效果的客观评价。通过分析所述合成图片中人像元素与衣物元素的匹配度，给出客观评价。首先需要建立分析策略，如各肤色与各衣物颜色的匹配度、各体型与各衣物种类、版型、尺寸的匹配度等，以暗黄肤色为例，其对白色的匹配度比对暗紫色的匹配度高。其次，对合成图片进行解析，再结合分析策略，给出所述合成图片中人像元素与衣物元素的匹配度评价，此匹配度为第一匹配度参数。
衣物图片推送模块，在形成第一匹配度参数后，结合用户的肤色、体型以及目标衣物图片中衣物元素的种类给用户推送匹配度高于第一匹配度参数的衣物图片。
本领域技术人员在考虑说明书及实践这里公开的发明后，将容易想到本公开的其它实施方案。本公开旨在涵盖本公开的任何变型、用途或者适应性变化，这些变型、用途或者适应性变化遵循本公开的一般性原理并包括本公开未公开的本技术领域中的公知常识或惯用技术手段。说明书和实施例仅被视为示例性的，本公开的真正范围和精神由权利要求指出。
应当注意的是，本发明的实施例有较佳的实施性，且并非对本发明作任何形式的限制，任何熟悉该领域的技术人员可能利用上述揭示的技术内容变更或修饰为等同的有效实施例，但凡未脱离本发明技术方案的内容，依据本发明的技术实质对以上实施例所作的任何修改或等同变化及修饰，均仍属于本发明技术方案的范围内。

Claims (10)

  1. 一种基于智能终端的试衣方法，其特征在于，包括以下步骤：
    获取包括人像元素的人像图片；
    获取包括衣物元素的目标衣物图片；
    将所述人像图片与所述目标衣物图片合成形成一合成图片，使所述人像图片中的人像元素穿戴有所述目标衣物图片中的衣物元素；
    显示所述合成图片。
  2. 如权利要求1所述的试衣方法，其特征在于，
    获取包括人像元素的人像图片的方法包括：
    通过照相机拍照获取人像图片，或获取所述智能终端内应用程序或图库内的人像图片；当获取人像图片的方法为通过照相机拍照获取时，其步骤包括：
    获取智能终端内的照相机权限或于智能终端内建立照相机模块；
    读取照相机拍取的人像图片；
    当获取人像图片的方法为获取所述智能终端内应用程序或图库内的人像图片时，其步骤包括：
    获取所述智能终端内应用程序或图库的访问权限；
    读取所述智能终端内应用程序或图库内的人像图片。
  3. 如权利要求1所述的试衣方法，其特征在于，
    获取包括衣物元素的目标衣物图片的方法包括：
    通过照相机拍照获取衣物图片、或获取所述智能终端内应用程序或图库内的衣物图片；当获取衣物图片的方法为通过照相机拍照获取时，其步骤包括：
    获取智能终端内的照相机权限或于智能终端内建立照相机模块；
    读取照相机拍取的衣物图片；
    当获取衣物图片的方法为获取所述智能终端内应用程序或图库内的衣物图片时，其步骤包括：
    获取所述智能终端内应用程序或图库的访问权限；
    读取所述智能终端内应用程序或图库内的衣物图片。
  4. 如权利要求1所述的试衣方法，其特征在于，
    将所述人像图片与所述目标衣物图片合成形成一合成图片，使所述人像图片中的人像元素穿戴有所述目标衣物图片中的衣物元素的步骤包括：
    捕捉所述人像图片中人像元素内的未被衣物覆盖部分；
    根据所述未被衣物覆盖部分重构所述人像元素内的衣物覆盖部分的轮廓；
    将所述衣物元素覆盖于所述衣物覆盖部分上。
  5. 如权利要求1所述的试衣方法，其特征在于，在显示所述合成图片的步骤后，所述方法还包括：
    分析所述合成图片中人像元素与衣物元素的匹配度，形成一第一匹配度参数；
    基于所述第一匹配度参数，推送一具有第二匹配度参数的衣物图片，其中所述第二匹配度参数大于所述第一匹配度参数。
  6. 一种基于智能终端的试衣系统，其特征在于，包括：
    人像图片获取模块，获取包含人像元素的人像图片；
    衣物图片获取模块，获取包含衣物元素的目标衣物图片；
    合成模块，将所述人像图片与所述目标衣物图片合成形成一合成图片，使所述人像图片中的人像元素穿戴有所述目标衣物图片中的衣物元素；
    显示模块，显示所述合成图片。
  7. 如权利要求6所述的试衣系统，其特征在于，
    所述人像图片获取模块包括：
    照相机建立单元，调用智能终端摄像头捕获并固定图片；
    照片读取单元，读取照相机内的人像图片；或
    所述人像图片获取模块包括：
    读取权限获取单元，向所述智能终端发送一权限请求，以获取所述智能终端内应用程序、图库读取权限；
    照片读取单元，读取应用程序或图库内的人像图片。
  8. 如权利要求6所述的试衣系统，其特征在于，
    所述衣物图片获取模块包括：
    照相机建立单元，调用智能终端摄像头捕获并固定图片；
    照片读取单元，读取照相机内的衣物图片；或
    所述衣物图片获取模块包括：
    读取权限获取单元，向所述智能终端发送一权限请求，以获取所述智能终端内应用程序、图库读取权限；
    照片读取单元，读取应用程序或图库内的衣物图片。
  9. 如权利要求6所述的试衣系统，其特征在于，
    所述合成模块包括：
    人像元素捕捉单元，捕捉所述人像图片中人像元素内的未被衣物覆盖部分；
    人像轮廓重构单元，根据所述未被衣物覆盖部分重构所述人像元素内的衣物覆盖部分的轮廓；
    覆盖单元，将所述衣物元素覆盖于所述衣物覆盖部分上。
  10. 如权利要求6所述的试衣系统，其特征在于，在显示模块后，所述系统还包括：
    匹配度分析模块，分析所述合成图片中人像元素与衣物元素的匹配度，形成一第一匹配度参数；
    衣物图片推送模块，基于所述第一匹配度参数，推送一具有第二匹配度参数的衣物图片，其中所述第二匹配度参数大于所述第一匹配度参数。
PCT/CN2017/095582 2017-08-02 2017-08-02 一种基于智能终端的试衣方法及试衣系统 Ceased WO2019023965A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/CN2017/095582 WO2019023965A1 (zh) 2017-08-02 2017-08-02 一种基于智能终端的试衣方法及试衣系统

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2017/095582 WO2019023965A1 (zh) 2017-08-02 2017-08-02 一种基于智能终端的试衣方法及试衣系统

Publications (1)

Publication Number Publication Date
WO2019023965A1 true WO2019023965A1 (zh) 2019-02-07

Family

ID=65232327

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2017/095582 Ceased WO2019023965A1 (zh) 2017-08-02 2017-08-02 一种基于智能终端的试衣方法及试衣系统

Country Status (1)

Country Link
WO (1) WO2019023965A1 (zh)

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1632832A (zh) * 2003-12-24 2005-06-29 毛新 在本人图像上进行网络试衣
CN104820930A (zh) * 2015-05-19 2015-08-05 北京五悦信息科技有限公司 一种智能试衣系统及其虚拟试衣方法
CN105608238A (zh) * 2014-11-21 2016-05-25 中兴通讯股份有限公司 服装试穿方法及装置
CN105812754A (zh) * 2016-05-31 2016-07-27 珠海市魅族科技有限公司 一种试装方法以及试装装置
US20170018024A1 (en) * 2015-07-15 2017-01-19 Futurewei Technologies, Inc. System and Method for Virtual Clothes Fitting Based on Video Augmented Reality in Mobile Phone

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN1632832A (zh) * 2003-12-24 2005-06-29 毛新 在本人图像上进行网络试衣
CN105608238A (zh) * 2014-11-21 2016-05-25 中兴通讯股份有限公司 服装试穿方法及装置
CN104820930A (zh) * 2015-05-19 2015-08-05 北京五悦信息科技有限公司 一种智能试衣系统及其虚拟试衣方法
US20170018024A1 (en) * 2015-07-15 2017-01-19 Futurewei Technologies, Inc. System and Method for Virtual Clothes Fitting Based on Video Augmented Reality in Mobile Phone
CN105812754A (zh) * 2016-05-31 2016-07-27 珠海市魅族科技有限公司 一种试装方法以及试装装置

Similar Documents

Publication Publication Date Title
US12062114B2 (en) Method, system, and device of virtual dressing utilizing image processing, machine learning, and computer vision
EP3479296B1 (en) System of virtual dressing utilizing image processing, machine learning, and computer vision
CN111681070B (zh) 线上商品的购买方法、购买装置、存储装置及购买设备
CN109325988B (zh) 一种面部表情合成方法、装置及电子设备
US10109315B2 (en) Devices, systems and methods for auto-delay video presentation
TWI227444B (en) Simulation method for make-up trial and the device thereof
US10311508B2 (en) Garment modeling simulation system and process
CN107665238B (zh) 图片处理方法和装置、用于图片处理的装置
CN104952113B (zh) 服饰试穿体验方法、系统及设备
CN110021061A (zh) 搭配模型构建方法、服饰推荐方法、装置、介质及终端
JP2019510297A (ja) ユーザの真実の人体モデルへの仮想的な試着
JP2010507854A (ja) ビデオイメージシーケンスを仮想シミュレーションする方法及び装置
JP7228025B2 (ja) 多重検出を伴う拡張現実に基づく衣服の仮想試着のための方法およびデバイス
CN105426462A (zh) 一种基于图像元素的图像搜索方法和装置
CN102546945A (zh) 手机拍照自动优化的方法
Thalmann et al. Modeling of populations
CN106373086A (zh) 一种基于增强现实的戒指2d在线实时试戴方法及系统
WO2020151427A1 (zh) 基于三维人体建模的智能制衣系统及方法
CN111182196A (zh) 拍照预览方法、智能终端及具有存储功能的装置
US10007860B1 (en) Identifying items in images using regions-of-interest
CN107665482A (zh) 实现双重曝光的视频数据实时处理方法及装置、计算设备
CN106855987A (zh) 基于模特道具的真实感时装展示方法和装置
WO2019090502A1 (zh) 一种基于智能终端的图像拍摄方法及图像拍摄系统
JP6544970B2 (ja) 画像処理装置、画像処理方法、及びプログラム
CN118396857A (zh) 一种图像处理方法及电子设备

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 17919887

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 17919887

Country of ref document: EP

Kind code of ref document: A1