WO2010010342A1 - Système et procédé de reconnaissance faciale - Google Patents

Système et procédé de reconnaissance faciale Download PDF

Info

Publication number
WO2010010342A1
WO2010010342A1 PCT/GB2009/001811 GB2009001811W WO2010010342A1 WO 2010010342 A1 WO2010010342 A1 WO 2010010342A1 GB 2009001811 W GB2009001811 W GB 2009001811W WO 2010010342 A1 WO2010010342 A1 WO 2010010342A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
face
image data
subject
model
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/GB2009/001811
Other languages
English (en)
Inventor
Josef Kittler
Xuan ZOU
Jose Rafael Tena
Peter Hancock
Alexandra Helen Mcintyre
Kieron Messer
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.)
Omniperception Ltd
Original Assignee
Omniperception 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 Omniperception Ltd filed Critical Omniperception Ltd
Publication of WO2010010342A1 publication Critical patent/WO2010010342A1/fr
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
    • G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
    • G06V40/16—Human faces, e.g. facial parts, sketches or expressions
    • G06V40/168—Feature extraction; Face representation

Definitions

  • This invention relates to a method and system for facial recognition of a subject's face and the related computer program product.
  • Human judgement is also disproportionally affected by superficial and external distractions such as a difference in face image colour tinge, in pose, or hairstyle, for example.
  • human operators are requested to identify a query face from a subset of candidate' s faces selected either manually or by an automatic face recognition system, by their very definition the subset of candidate's faces will tend to be similar. The operator is then more likely to be confused by the apparent similarity among these candidates' faces.
  • a method of human recognition of a subject's face comprising the steps of producing image data representing a morphable image of the subject's face; processing said image data to produce a morphed image of the subject's face and presenting said morphed image for recognition.
  • a facial image processing system for presenting an image of the subject's face for recognition of said subject comprising, acquisition means for acquiring image data representing a morphable image of the subject's face; processing means for processing said image data to produce a morphed image of the subjects face; presentation means for presenting said morphed image for recognition of said subject.
  • a computer program product for use in facial recognition of a subject's face comprising; machine readable media having recorded thereon programming instructions for causing a computer system to: acquire image data representing; a morphable image of the subject' face; process said image data to produce a morphed image of the subject's face; and present said morphed image for recognition.
  • the method, system and computer program product to be described herein focus on applying image morphing techniques to acquired face images of a subject to help recognition of the acquired image by a human. This is generally done by processing image data obtained from the face image of the subject After caricaturing of the image (caricaturing is a specific morphing technique), a face can generally be more easily recognised by a human. Other morphing techniques also make a face easier to recognise.
  • the image morphing technique developed for this system is based on the 3D face morphing process.
  • the technique involves fitting a generic 3D face morphable model to image data representing a 2D face image of a subject to obtain image data representing a morphable 3D face of the image of the subject, which can then be processed by morphing on the shape channel and the texture channel, either separately or together.
  • the 3D face morphing process facilitates 3D face caricaturing, and this provides considerably more information than 2D face morphing for a human operator, in the sense that a face with a new pose, under different illumination and improved resolution can be generated when the 3D information is available. This change in pose, resolution and illumination is generally not available if the image of the subject is morphed by a 2D process only.
  • Figure 1 shows an example of a human recognition task
  • Figure 2 shows an example of a human recognition task with the subject in an alternative pose
  • Figure 3 shows 3D face morphing along a line connecting an average face and a specific face
  • Figure 4 shows a statistical shape model and statistical texture model of a 3D face
  • Figure 5 is a diagram of a 3D morphing system
  • Figure 6 shows fitting a 3D morphable model to a 2D face to obtain a 3D face of the subject
  • Figure 7 shows morphing from a male face to a female face on the shape and texture channel
  • Figure 8 shows morphing a male face to a female face on the shape channel only
  • Figure 9 shows morphing a male face to a female face on the texture channel only
  • Figure 10 shows morphing the average face by varying the first coefficient of the shape model but retaining the texture coefficient
  • Figure 11 shows morphing the average face in the texture channel along the direction defined by two female faces with different skin colour.
  • Figure 1 shows an example of a human recognition task. The task is to match the query face (10) in the top row with one of the ten possible candidates (2OA 20J) in the array of gallery images below. More details of this particular task are described in Vicki Bruce, Zo Henderson, Karen Greenwood, Peter J. B. Hancock, A. Mike Burton, and Paul Miller. Verification of face identities from images captured on video. Journal of Experimental Psychology: Applied, 5:339-360,1999.
  • Figure 2 shows a recognition task where the subject to be identified (12) is given an unfamiliar pose. Again, the subject should be matched with one of ten candidate faces (2OA 20J) in the array of gallery images below. In this case recognition of the face by a human is generally more difficult, due to the unfamiliar pose of the subject to be identified.
  • the image data can be processed in such a way that the image of the subject to be recognised undergoes 2D image morphing and/or by 3D image morphing. Different types of image morphing will now be described.
  • Image Morphing is defined as the process in which one image changes to another through a seamless transition. This applies to many different sort of images, and not just faces.
  • the traditional technique for 2D image morphing has matured over the last decade.
  • 3D image morphing has several advantages over 2D image morphing.
  • it is easy to change the view point of the face, the particular facial illumination conditions and to deal with any facial occlusions that may be present in the 2D image.
  • the surface of a particular 3D object is described by a 3D shape map and a 2D texture map of the surface.
  • the 3D shape map contains information about the 3D form of the face and is represented in terms of a mesh of vertices defined by their 3D coordinates, and how several vertices of the mesh form each cell of the mesh.
  • the texture map is a colour image that captures the colour information of the surface.
  • Morphing from one 3D face to another 3D face can then be performed on the shape channel only, on the texture channel only, or simultaneously, on both the shape and texture channels.
  • the morphing between these two surfaces can be performed as the interpolation between the coordinates of the corresponding vertices on each surface.
  • the morphing on the 2D texture map is implemented by the interpolation between the corresponding pixel locations in the source images, which is just the same as in the Cross-dissolves method for 2D image morphing (The Cross- dissolves method is described in detail in the Wolberg reference mentioned previously.
  • Morphing between two topologically different 3D object surfaces is an interesting but challenging process.
  • 3D correspondence between the source and the target 3D surface has to be established by surface registration. http://web. mit. edu/manoli/morph/www/morph. html, describes an example of morphing between two topologically different 3D object surfaces which involves mesh subdivision, union, and matching.
  • 3D animation software systems there are some software systems specifically designed for 3D face animation, such as Facial Studio and FaceGen Modeler (http://www. face sen, cow).
  • FaceGen Modeller is able to generate various face effects based on a statistical 3D face model, and the software also has a function for 3D reconstruction of a face from a single 2D face image or from an orthogonal 2D face image pair.
  • these 3D animation software systems are developed specifically for face animation and cannot be applied directly for the purpose of assisting face matching by human operators.
  • An alternative image morphing process is that of face caricaturing. This is the process of generating a face caricature in which distinctive facial features are exaggerated. Previously face caricatures could only be created by an artist, but nowadays caricatures can be generated automatically by computers. Image morphing provides the means to perform face caricaturing.
  • a face is a point in a high dimensional space.
  • Mathematically the linear morphing along a line defined by two faces can be represented by:
  • M is the resulting morphed face
  • S is a test average face from a set of reference images
  • D is the face of the subject to be recognised
  • ⁇ is the morphing coefficient.
  • the reference images are typically obtained by processing (in various different ways) a set of training images. However, they may be obtained without the use of training images.
  • S is the average face
  • M is a caricature of face D when ⁇ > 1.
  • the distinctive facial features in face D become more exaggerated in M when ⁇ gets larger.
  • the morphing generates so-called anti-faces. Between the average face and the given face (1 > ⁇ > 0), the morphing is called anti- caricaturing. This is described in more detail in D.
  • FIG. 3 An example of a 3D face morphing process is shown in Fig. 3.
  • This Figure shows a range of faces for various values of ⁇ .
  • the third face along from the left hand side is an average 3D face (40)
  • the third face from the right hand side (30) is a real face and between the average face and the real face the real face has undergone a series of anti-caricaturing steps to produce several intermediate faces (30A, 30B, 30C).
  • the two faces (30AA, 60) on the right hand side of the Figure are caricatures of the real 3D face (30) in Figure 3.
  • the first automatic caricaturing system was proposed in S. E. Brennan. Caricature generator: the dynamic exaggeration of faces by computer. Leonardo, 18(3): 170- 178, 1985, to generate 2D line drawn caricatures from photographs.
  • Other caricaturing systems and related image processing systems are described in Philip J. Benson and David I. Perrett. Synthesizing continuous-tone caricatures. Image Vision Comput., 9(2): 123-129, 1991, and D. A. Rowland, D. I. Perrett, D. M. Burt, K J. Lee, and S. Kamatsu. Transforming facial images in 2 and 3D. In Imagina 97 -conference- ACES Proceedings, pages 159-175, 1997.
  • Equation 1 The linear morphing described by equation 1 can be generalised to morphing a third face along any direction defined by two example faces.
  • Sl and S2 are two images from the set of reference images defining a morphing direction.
  • the resulting image M will represent the subject's face D, subject to the given holistic variation.
  • Sl and S2 are the average male face and the average female face, respectively, the morphing will carry changes related to gender change.
  • Sl and S2 may represent average faces of different race, or of different ages, or any other identifiable cohort.
  • An alternative morphing process is statistical model based morphing. This is a method which makes use of the statistical information about an object class conveyed by its image.
  • the source image and the target images are images of objects in the same class, e.g. both images are faces, the morphing can be performed with the statistical model of the class.
  • the images or data representing the images are projected to a subspace spanned by a statistical model, e.g. Principle Component Analysis (PCA) subspace, and each image or image data is represented by projection coefficients.
  • a statistical model e.g. Principle Component Analysis (PCA) subspace
  • Morphing can then be performed by interpolation between the projection coefficients of the source image (or image data) and those of the target image (or image data).
  • Each morphed image can then be reconstructed from the corresponding projection coefficients obtained by the interpolation process.
  • the projection coefficients have been described here within reference to statistical model based morphing they may also be derived from all the other morphing processes described herein.
  • a statistical model of a 3D face usually consists of a statistical shape model and a statistical texture model, this is illustrated in more detail in Figure 4.
  • the 3D face (70) to be modelled is shown on the left hand side of the Figure.
  • Shape models (72a, 72b 72n) are shown in the top row, alongside references A 1 , A2 etc.
  • 3D face images e.g. photographs
  • 3D face data is not as readily available as 2D face images.
  • a 3D model fitting technique can be applied to reconstruct a 3D face from the given 2D face images, then 3D morphing can be performed on the reconstructed 3D face.
  • the significance of the 3D model fitting technique is that a face in a familiar view angle can be rendered by fitting the model to face in any arbitrary view, which is definitely helpful to human face recognition.
  • the target image (12) shown in Figure 2 could be modelled and the view of the face could be changed to be a direct 'front on' view, rather than a face looking to the left.
  • the initial work on 3D morphable model is described in V. Blanz and T. Vetter. A morphable model for the synthesis of 3D faces. In Proceedings of SIGGRAPH, 1999.
  • the morphable model is a statistical face model which consists of a PCA model of face shape and a PCA model of face texture map.
  • the morphable model is trained on 3D faces after they are captured by laser scanner Cyberware and registered with a method based on optical flow.
  • FIG. 5 A flow diagram of a proposed 3D face morphing system is shown in Figure 5.
  • the system typically includes images capture means such as a camera (not shown), and image processing means (not shown), which is generally a computer system.
  • the face morphing system takes a 2D face image (80) of a subject to be recognised as an input image and reconstructs a corresponding 3D face of the recognised and reconstructs a corresponding 3D face by fitting (82) a generic 3D morphable model (84) to the 2D face image.
  • the 2D face image (80) of the subject' face is typically obtained using the image capture means.
  • the 2D image is then processed by the image processing means to generate image data (for example, pixel intensity) which can undergo further processing.
  • image data for example, pixel intensity
  • the initial processing of the 2D image to generate the image data may take place in the camera or in the associated computer system.
  • the subsequent processing steps will generally take place in the computer system, which has been appropriately programmed to perform the processing steps.
  • the image data generated from the 2D face image obtained by the camera represents a morphable image (86) of the subject's face.
  • the subsequent processing of the image data will morph the morphable image (86) to produce a morphed image (96) of the subject's face.
  • data representing a generic 3D morphable model (84) is fitted to the image data representing the 2D face image (80) to produce image data representing a 3D morphable face (86).
  • the image data representing the 3D morphable face (86) can be subsequently processed to ultimately generate a final morphed image (96) of the subject's face.
  • this image data can be performed, each way is equivalent to morphing the 3D morphable face (86) in an alternative way.
  • the particular processing performed on the image data representing the 3D morphable face is generally based on the choice of the operator of the 3D face morphing system.
  • Four different types of morphing are shown in Figure 5. These are 3D caricaturing (88), morphing on a defined direction (90), multiple channel morphing (92) and model based morphing (94), all of which have been described above.
  • the result of the morphing is a morphed face image (96).
  • the resultant morphed image (96) is then easier for another human to identify. AU of these steps can be performed on a standard computer system loaded with appropriate software for performing the various steps required.
  • the 3D face morphing system described with reference to figure 5 uses a texture map warped to the 3D shape model. This allows texture detail of the face to be preserved, despite using a 3D shape model of only 6000 polygons.
  • the 3D morphable model can be represented by projection coefficients.
  • an optimisation process is needed to tune the shape coefficients and texture coefficients along with rendering parameters to minimise the difference of the input image and the rendered image based on those optimised shape and texture coefficients.
  • the rendering parameters include pose angles, 3D translation, ambient light intensities, directed light intensities and angles, and other parameters of the camera and colour channels.
  • the well known illumination model of Phong (B. T. Phong, "Illumination for Computer Generated Pictures” Comm ACM VoI 18(6) p. 311 - 317 June 1975) is adopted in the rendering process to describe the diffuse and specular reflection of the surface.
  • the various functionalities of the 3D face morphing system described herein with reference to figure 5 include 3D model fitting, 3D caricaturing, multiple channel morphing, model based morphing, and morphing on a user-defined direction.
  • Figure 6 shows a 3D face reconstructed from a 2D image (100) using a morphable model (84).
  • the 2D image (100) on the left hand side of the figure is the original 2D facial image of the subject to be recognised. This is fitted
  • Image data representing the 2D image is obtained as described above with reference to Figure 5.
  • the obtained data is then processed by fitting data (82) representing the 3D morphable face (84) to the image data of the 2D image.
  • This model fitting stage will result in the generation of a 3D morphable face.
  • the right hand side of the figure shows example 2D images (96 A, 96B, 96C) of the resultant 3D face from various different directions.
  • the original subject face to be recognised (100) can be rendered in various different, arbitrary poses.
  • the subject's face can be rendered to appear 'front on', instead of slightly posed to the right hand side
  • a 'front on' image is generally easier for a human to recognise than an image that is not 'front on' .
  • 3D caricaturing is implemented as a morphing along the line connecting the average 3D face obtained from a set of training images and the given face to be recognised. The whole process is illustrated in Figure 3 as described earlier.
  • Image data representing the resultant 3D face (30) is then processed according to the caricature process described earlier.
  • the facial features to be caricatured are selected by the operator who will use their judgement in view of the overall impression created by the subject's face. For example, the operator may decide to create a caricature by changing (increasing/reducing) the size of a facial feature, or by changing the intensity of the colour of a facial feature (e.g. heighten facial redness, or increase the intensity of eye colour). Other type of caricature (not specifically described) may also be performed.
  • the operator can also instruct the system to perform a series of caricatures, rather than being limited to just changing one facial feature. For example, the operator may decide that they wish to increase the size of the subject's nose, then to reduce the size of the subject's ears, then to make the chin more pronounced, and so on.
  • Figure 7 illustrates the change in a subject's face as the image data is processed so that the 3D morphable face is morphed along the shape and texture channel from a male face (1 10) to a female face (1 12).
  • This figure shows 6 different intermediate faces (HOA, HOB .7) as the 3D morphable face is morphed from male to female, but there may be any number of intermediate faces between the male (110) and female faces (112).
  • Figure 8 also illustrates the change in a face as it is morphed along the shape channel only from a male face (114) to a female face (116). Again, the morphing arises as a result of processing the image data of the 3D morphable face. No changes occur on the texture channel, this means only that shape of the face is changed, there are no changes to the texture of the face.
  • this Figure shows six different intermediate faces (114 A, 114B ) in the change from male (114) to female (116), but again there may be any number of intermediate faces (114A 114N) between the starting male face (114) and the final female face (116).
  • Figure 9 illustrates the change in a face as it is morphed along the texture channel only from a male face (118) to a female face (120). Again, the morphing arises as a result of processing the image data of the 3D morphable face to morph the 3D morphable face. In this embodiment no facial changes occur on the shape channel and so the shape of the face is unchanged. As for Figures 7 and 8 there may be any number of intermediate faces (1 18A ....118N) between the starting face (118) and the final face (120).
  • the starting face (110, 114, 118) is male and the end face is female (112, 1 16, 120).
  • the starting face may be female and be transformed by processing the image data and morphing to a male face (again with any number of intermediate faces), or the gender of the starting face (male/female) may be unchanged by the morphing process as the image data is processed and morphing on shape and/or texture channels may simply change the appearance of the face but not the gender of the subject.
  • all of the images in the morphing process are shown in frontal view.
  • the faces may be shown within a particular pose (non-frontal view) and/or with a particular tilt of the face.
  • the image data can be processed in such a way that morphing can be carried out on the model coefficients, and the desired faces can be reconstructed from the resulting projection coefficients after the processing has been completed.
  • the result will generally exhibit no difference from processing image data of the
  • Figure 10 illustrates the change in an average face (122) resulting from model based morphing by varying a first coefficient of the image data of the shape model, but leaving the coefficients of the image data texture model unchanged.
  • the face after the processing of the coefficients is at the right hand side (124).
  • the gender of the subject has not changed ( Figure 10 shows an average male face, but the processing of the image data and subsequently the morphing could apply to a female face), but this technique can be applied to change the gender from male to female and vice versa.
  • model based morphing could also be used to vary coefficients of the texture model, whilst leaving the shape model unchanged, or may vary coefficients of both the shape and texture model together. It seems that this first coefficient of the shape controls the ratio model of the size of the upper face to that of the lower part. Also there may be any number of intermediate faces (122 A....122N) between the initial face (122) and the final face (124). 5. Morphing along defined directions:
  • Another function provided by the system is processing the image data so that the 3D morphable face is morphed along a direction defined by two faces (126, 128) selected by the user.
  • An example of processing the image data to morph the average face in the texture channel along the direction defined by two female faces with different skin colour is shown in Figure 11.
  • the morphing induces a transform of skin colour from one subject (130) to the other (132).
  • the starting average female face (130) is European, and the final face (132) is an Afro-
  • Carribean face The face passes through intermediate faces (130A, 130B ) as the morphing proceeds along the defined direction.
  • the morphing may be along a texture channel and/or along a shape channel of the defined direction.
  • the morphing generally will change one or more race specific characteristics of the morphable image, for example, the characteristic may be skin colour, hair colour, hair texture, facial feature size and facial feature orientation.
  • the gender of the face is not changed, but the morphing process may be such as to change the gender of the starting subject, as well as changing at least one race specific characteristic.
  • the race of the two faces defining the direction may be different, e.g. European, African, Indian Asian, Chinese Asian.
  • the skin colour may not change but other race characteristics will change.
  • the reference images that define the morphing direction will be pre-set.
  • the system may include various pre-defined directions related to race or gender morphing, for example. In this case, the operator will select the particular morphing operation to be performed on the image of the subject's face and the relevant reference images defining the necessary direction will be selected from a central database.
  • the operator may decide to manually review the images in the central database and to select particular reference images to define the morphing direction for a particular morphing operation to be carried out.
  • Figure 5 shows a two stage morphing system. That is, a model fitting stage (82) to produce a 3D morphable face (86), followed by four alternative methods (88, 90, 92, 94) of morphing the resultant morphable 3D face.
  • the various morphing methods have all be described above.
  • a 3D face model (80) that has been morphed by one of the four morphing processes (88, 90, 92, 94) to undergo one or more further morphing steps.
  • the subsequent morphing steps may be carried out by performing any of the four types of morphing process. It does not have to be the same as the morphing process initially performed, although the same morphing techniques may be used.
  • the image data could be morphed on multiple channels (shape/texture) to change the gender of the subject, and the resultant morphed image may be morphed again on multiple channels to change a particular facial shape/texture characteristic.
  • the image data could be morphed on multiple channels to change the gender of the subject, and the resultant morphed image may be morphed again, by a 3D caricature process, to change the size of a particular facial feature, for example.
  • the resultant morphed image from a multiple (2 stage) morphing process may then be present to the operator for recognition or they may undergo yet more morphing.
  • the morphed image of the subject's face can be compared to one or more images from a pre-existing set of gallery images of target faces.
  • the gallery images are stored on the computer system, and a subset of gallery images with facial features similar to the subject can be called up.
  • the gallery images selected from comparison will also have all of these specific features.
  • the gallery is generally provided with enough gallery images to produce subset of gallery images for all races, ages, gender and other facial characteristics.
  • the subset of gallery images can be selected manually by the operator after review of the entire gallery, or automatically by the system, once various parameters (race, hair colour, age, gender, etc ..) have been entered by the system operator to define the facial features of the subject. However, the subset of gallery images is selected they are then displayed, on a computer screen for example, for the human operator to review.
  • the human operator will look at the morphed image of the subject, and compare the image with the selected gallery images and then decide which of the gallery images corresponds to an image of the subject.
  • the various different morphing processes described herein can also be applied to the gallery images.
  • one or more of the gallery images can be morphed to produce morphed gallery images.
  • the morphed gallery images can be compared with an unmorphed directly obtained image of the subject, to assist in recognition of the subject by a human operator.
  • the image of the subject's face is morphed as described previously, and one or more of the gallery images can also be morphed.
  • the images will undergo the same type of morphing in the same sequence e.g. change of gender, then change of race, then caricature of a particular facial feature.
  • the subject's face undergoes one or more morphing steps e.g. change of race, then change of gender, then change of eye colour, whereas the gallery images undergo more or less morphing steps e.g. only the race of the gallery image is changed, and vice versa.
  • the resultant morphed image of the subject, and morphed gallery images can then be presented, on a computer screen for example, to the human operator to assist recognition of the subject's face by the operator.
  • the faces represented by each of the gallery images will undergo the same model based morphing as the image of the subject's face. This will make it easier for the operator to identify the subject's face as one of the faces presented in the gallery images.
  • a morphed face it is generally easier for a morphed face to be recognised if the race of the image of the subject's face is morphed to be the same as the race of the operator who is required to identify the image of the subject.

Landscapes

  • Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Oral & Maxillofacial Surgery (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Health & Medical Sciences (AREA)
  • Human Computer Interaction (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Theoretical Computer Science (AREA)
  • Processing Or Creating Images (AREA)

Abstract

La présente invention porte sur un procédé de reconnaissance du visage d'un sujet. Le procédé traite des données d'image du visage d'un sujet, traite ensuite les données d'image pour produire une image transformée par morphage qui peut être présentée pour reconnaissance. L'invention porte également sur un système apparenté et sur un produit-programme informatique apparenté permettant une reconnaissance du visage d'un sujet.
PCT/GB2009/001811 2008-07-25 2009-07-21 Système et procédé de reconnaissance faciale Ceased WO2010010342A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
GBGB0813685.5A GB0813685D0 (en) 2008-07-25 2008-07-25 A system & method for facial recognition
GB0813685.5 2008-07-25

Publications (1)

Publication Number Publication Date
WO2010010342A1 true WO2010010342A1 (fr) 2010-01-28

Family

ID=39746969

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/GB2009/001811 Ceased WO2010010342A1 (fr) 2008-07-25 2009-07-21 Système et procédé de reconnaissance faciale

Country Status (2)

Country Link
GB (1) GB0813685D0 (fr)
WO (1) WO2010010342A1 (fr)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109416843A (zh) * 2016-05-13 2019-03-01 帝国科技及医学学院 实时高度映射
CN110069655A (zh) * 2019-03-26 2019-07-30 中山大学 一种私人相册的人脸搜索方法
US10379734B2 (en) 2013-02-23 2019-08-13 Qualcomm Incorporated Systems and methods for interactive image caricaturing by an electronic device

Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP1039417A1 (fr) * 1999-03-19 2000-09-27 Max-Planck-Gesellschaft zur Förderung der Wissenschaften e.V. Méthode et appareil de traitement d'images basés sur des modèles à métamorphose

Patent Citations (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP1039417A1 (fr) * 1999-03-19 2000-09-27 Max-Planck-Gesellschaft zur Förderung der Wissenschaften e.V. Méthode et appareil de traitement d'images basés sur des modèles à métamorphose

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
COSTEN N ET AL: "Manifold caricatures: on the psychological consistency of computer face recognition", AUTOMATIC FACE AND GESTURE RECOGNITION, 1996., PROCEEDINGS OF THE SECO ND INTERNATIONAL CONFERENCE ON KILLINGTON, VT, USA 14-16 OCT. 1996, LOS ALAMITOS, CA, USA,IEEE COMPUT. SOC, US, 14 October 1996 (1996-10-14), pages 4 - 9, XP010200392, ISBN: 978-0-8186-7713-7 *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US10379734B2 (en) 2013-02-23 2019-08-13 Qualcomm Incorporated Systems and methods for interactive image caricaturing by an electronic device
US10761721B2 (en) 2013-02-23 2020-09-01 Qualcomm Incorporated Systems and methods for interactive image caricaturing by an electronic device
US11526272B2 (en) 2013-02-23 2022-12-13 Qualcomm Incorporated Systems and methods for interactive image caricaturing by an electronic device
CN109416843A (zh) * 2016-05-13 2019-03-01 帝国科技及医学学院 实时高度映射
CN110069655A (zh) * 2019-03-26 2019-07-30 中山大学 一种私人相册的人脸搜索方法

Also Published As

Publication number Publication date
GB0813685D0 (en) 2008-09-03

Similar Documents

Publication Publication Date Title
Yamaguchi et al. High-fidelity facial reflectance and geometry inference from an unconstrained image
Tran et al. On learning 3d face morphable model from in-the-wild images
US11625878B2 (en) Method, apparatus, and system generating 3D avatar from 2D image
Sharma et al. 3d face reconstruction in deep learning era: A survey
O'Toole et al. 3D shape and 2D surface textures of human faces: The role of “averages” in attractiveness and age
Tran et al. Nonlinear 3d face morphable model
US10796480B2 (en) Methods of generating personalized 3D head models or 3D body models
CN108573527B (zh) 一种表情图片生成方法及其设备、存储介质
Jeni et al. Dense 3D face alignment from 2D video for real-time use
WO2022095721A1 (fr) Procédé et appareil de formation de modèle d'estimation de paramètre, dispositif, et support de stockage
Suwajanakorn et al. What makes tom hanks look like tom hanks
JP6207210B2 (ja) 情報処理装置およびその方法
US20080309662A1 (en) Example Based 3D Reconstruction
Kittler et al. 3D morphable face models and their applications
CN109690617A (zh) 用于数字化妆镜的系统和方法
CN116097319A (zh) 利用空间感知的条件gan的高分辨率可控面部老化
JP2023531265A5 (fr)
CN113628327A (zh) 一种头部三维重建方法及设备
Geng et al. Towards photo-realistic facial expression manipulation
Ren et al. Make-a-character: High quality text-to-3d character generation within minutes
Qammaz et al. A unified approach for occlusion tolerant 3d facial pose capture and gaze estimation using mocapnets
Rashid et al. High-fidelity facial expression transfer using part-based local–global conditional gans
WO2010010342A1 (fr) Système et procédé de reconnaissance faciale
Ding et al. 3D face sparse reconstruction based on local linear fitting
Zhang et al. Monocular face reconstruction with global and local shape constraints

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: 09784763

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: 09784763

Country of ref document: EP

Kind code of ref document: A1