JP2009294958A - Method of synthesizing face image - Google Patents
Method of synthesizing face image Download PDFInfo
- Publication number
- JP2009294958A JP2009294958A JP2008148648A JP2008148648A JP2009294958A JP 2009294958 A JP2009294958 A JP 2009294958A JP 2008148648 A JP2008148648 A JP 2008148648A JP 2008148648 A JP2008148648 A JP 2008148648A JP 2009294958 A JP2009294958 A JP 2009294958A
- Authority
- JP
- Japan
- Prior art keywords
- face
- image
- standardized
- images
- original image
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 33
- 230000002194 synthesizing effect Effects 0.000 title claims abstract description 11
- 238000000513 principal component analysis Methods 0.000 claims abstract description 23
- 210000001508 eye Anatomy 0.000 claims abstract description 12
- 238000004364 calculation method Methods 0.000 claims description 12
- 239000000203 mixture Substances 0.000 claims description 11
- 239000003086 colorant Substances 0.000 claims description 9
- 210000004709 eyebrow Anatomy 0.000 claims description 9
- 230000001815 facial effect Effects 0.000 claims description 9
- 210000004209 hair Anatomy 0.000 claims description 8
- 238000012935 Averaging Methods 0.000 claims description 6
- 239000002131 composite material Substances 0.000 claims description 6
- 238000000611 regression analysis Methods 0.000 claims description 6
- 230000000007 visual effect Effects 0.000 claims description 5
- 210000000214 mouth Anatomy 0.000 claims description 4
- 238000001308 synthesis method Methods 0.000 claims description 4
- 238000013461 design Methods 0.000 abstract description 4
- 239000002537 cosmetic Substances 0.000 description 7
- 238000012545 processing Methods 0.000 description 6
- 230000015572 biosynthetic process Effects 0.000 description 4
- 230000000873 masking effect Effects 0.000 description 4
- 238000010606 normalization Methods 0.000 description 4
- 238000003786 synthesis reaction Methods 0.000 description 4
- 230000003796 beauty Effects 0.000 description 3
- 230000007423 decrease Effects 0.000 description 3
- 210000000887 face Anatomy 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000004891 communication Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 210000000744 eyelid Anatomy 0.000 description 2
- 210000000088 lip Anatomy 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 230000001149 cognitive effect Effects 0.000 description 1
- 244000221110 common millet Species 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000012847 principal component analysis method Methods 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
Landscapes
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
Description
本発明は、見た目年齢が異なる顔画像を合成する顔画像の合成方法及び合成装置に関する。 The present invention relates to a face image composition method and composition apparatus for compositing face images having different appearance ages.
画像上で化粧のシミュレーションを行い、化粧方法や化粧品の選択のアドバイスを行う方法が開発されている（特許文献１）。この場合、シミュレーションの元画像には、モデルや顧客の顔画像が使用されている。 A method has been developed in which makeup simulation is performed on an image and advice on selection of a makeup method and cosmetics is made (Patent Document 1). In this case, a model or a customer's face image is used as a simulation original image.
一方、化粧の目的の一つに、実年齢よりも見た目年齢を低く又は高く見せることがある。そのために、所期の見た目年齢となるように、肌の色を調整したり陰影をつけたりすることがなされるが、美容部員が、輪郭、立体感、肌の色等が種々異なる個々の顧客に対し、要望通りの見た目年齢に化粧を施すことや、そのような化粧方法を顧客にアドバイスすることは容易でない。これに対し、輪郭等が異なる顧客ごとに見た目年齢の異なる顔画像を参照できれば、そのような化粧の完成度が高まり、顧客に対する化粧方法のアドバイスも適確となる。また、同一人について見た目年齢の異なる顔画像を得ることは、化粧品の研究、開発においても必要となる。 On the other hand, one of the purposes of makeup is to make the appearance age lower or higher than the actual age. For this purpose, the skin color is adjusted or shaded so as to achieve the desired appearance age, but the beauty staff can deal with individual customers with different outlines, three-dimensionality, skin color, etc. On the other hand, it is not easy to apply makeup to the appearance age as desired and to advise customers on such makeup methods. On the other hand, if facial images with different appearance ages can be referenced for each customer with different outlines, the degree of perfection of such makeup will increase, and advice on makeup methods for customers will be appropriate. It is also necessary for research and development of cosmetics to obtain facial images with different appearance ages for the same person.
そこで、本発明は、任意の顧客等の特定人の顔画像において見た目年齢が異なる顔画像を得られるようにすることを目的とする。 Therefore, an object of the present invention is to make it possible to obtain face images having different appearance ages in face images of a specific person such as an arbitrary customer.
本発明者は、
(a)複数の顔画像（以下、元画像という）から、その形状の特徴点が所定形状の顔型の特徴点に揃うように規格化した画像（以下、規格化元画像という）を合成し、その規格化元画像の色について主成分分析することにより固有ベクトル（以下、パラメータという）を得ると、パラメータの中には、見た目年齢に大きく寄与するものがあること、(b)規格化元画像について、前記主成分分析で得られる固有ベクトル空間における主成分スコア（以下、パラメータの重み係数という）のうち見た目年齢に寄与するものを変えた画像を得、それを元の形状に戻してやると、顔の輪郭や、眉、目、鼻、口の位置や形状等の顔の基本設計は元画像と共通だが、見た目年齢が異なる顔画像を得られることを見出した。
The inventor
(a) A composite image (hereinafter referred to as a normalized original image) is synthesized from a plurality of face images (hereinafter referred to as original images) so that the feature points of the shape are aligned with the face-shaped feature points of a predetermined shape. When eigenvectors (hereinafter referred to as parameters) are obtained by performing principal component analysis on the color of the standardized original image, some parameters contribute greatly to the appearance age, and (b) the standardized original image Is obtained by changing the principal component score in the eigenvector space obtained by the principal component analysis (hereinafter referred to as parameter weighting coefficient) that contributes to the apparent age, and returning it to its original shape, The basic design of the face, such as the outline and the position and shape of the eyebrows, eyes, nose, and mouth, is the same as the original image, but it was found that face images with different appearance ages can be obtained.
即ち、本発明は、見た目年齢の異なる顔画像を合成する方法であって、
複数人の元画像を取得し、
各元画像から規格化元画像を作成し、
複数人の規格化元画像の色について主成分分析を行い、
主成分分析により得られる規格化元画像のパラメータ中、見た目年齢に寄与するパラメータを特定し、そのパラメータの成分画像を作成し、
一方、任意の顔の規格化元画像を作成し、任意の顔の規格化元画像において、前記パラメータの重み係数を変化させることにより見た目年齢の異なる規格化顔画像を作成する顔画像の合成方法を提供し、また、特定人の顔の規格化元画像について見た目年齢の異なる規格か顔画像を作成した後、その特徴点を特定人の元画像の特徴点に戻すことにより、見た目年齢の異なる特定人の顔画像を作成する顔画像の合成方法を提供する。
That is, the present invention is a method for synthesizing face images having different appearance ages,
Get the original images of multiple people
Create a standardized original image from each original image,
Perform principal component analysis on the colors of the standardized original images of multiple people,
Among the parameters of the standardized original image obtained by principal component analysis, specify the parameter that contributes to the appearance age, create a component image of that parameter,
On the other hand, a method for synthesizing a face image that creates a standardized original image of an arbitrary face and creates a standardized face image having a different appearance age by changing a weighting coefficient of the parameter in the standardized original image of the arbitrary face In addition, after creating a standard or face image with a different appearance age for the standardized original image of a specific person's face, the feature point is changed back to the feature point of the original image of the specific person, so that the appearance age differs Provided is a face image synthesis method for creating a face image of a specific person.
また、本発明は、元画像を取得する画像取得手段と、元画像から見た目年齢の異なる合成顔画像を作成する演算手段とを備えた顔画像の合成装置であって、
該演算手段が、
元画像から規格化元画像を作成する機能、
複数人の規格化元画像の色について主成分分析を行う機能、
主成分分析による規格化元画像のパラメータ中、見た目年齢に寄与するパラメータを特定する機能、
任意の顔の規格化元画像において、前記パラメータの重み係数を変化させることにより、見た目年齢の異なる規格化顔画像を作成する機能を備えている顔画像の合成装置を提供し、さらに、見た目年齢の異なる規格化顔画像の特徴点を前記任意の顔の元画像の特徴点に戻すことにより、前記任意の顔について見た目年齢の異なる合成顔画像を形成する機能、
を備えている顔画像の合成装置を提供する。
Further, the present invention is a face image synthesizing device comprising an image acquisition means for acquiring an original image and a calculation means for creating a composite face image having a different apparent age from the original image,
The computing means is
A function to create a standardized original image from the original image,
A function to perform principal component analysis on the colors of the standardized original images of multiple people,
A function that identifies parameters that contribute to visual age among the parameters of the standardized original image by principal component analysis,
Provided is a face image synthesizing device having a function of creating a standardized face image having a different appearance age by changing a weighting factor of the parameter in a standardized original image of an arbitrary face, and further, an appearance age A function of forming a composite face image having different appearance ages with respect to the arbitrary face by returning the feature points of the standardized face image of different from the feature points of the original image of the arbitrary face,
An apparatus for synthesizing a face image is provided.
本発明によれば、顔の輪郭や、眉、目、鼻、口の位置や形状等の顔の基本設計は共通だが、色と陰影分布が異なることにより、見た目年齢の異なる顔画像、即ち、若く又は老いて見える顔画像を合成することができる。したがって、本発明により得られる顔画像を参照することにより、若く又は老けて見せる化粧方法、化粧品素材、化粧料等の開発や研究を推進させることができ、また、顧客を若く又は老けて見せるための化粧方法のアドバイスを適確に行うことが可能となる。 According to the present invention, the basic design of the face such as the outline of the face and the position and shape of the eyebrows, eyes, nose, mouth, etc. are common, but the facial images with different appearance ages due to the different color and shade distribution, i.e. A face image that looks young or old can be synthesized. Therefore, by referring to the face image obtained by the present invention, it is possible to promote the development and research of makeup methods, cosmetic materials, cosmetics, etc. that make you look younger or older, and to make customers look younger or older It is possible to give advice on makeup methods accurately.
以下、図面を参照しつつ、本発明を詳細に説明する。なお、各図中、同一符号は同一又は同等の構成要素を表している。 Hereinafter, the present invention will be described in detail with reference to the drawings. In each figure, the same numerals indicate the same or equivalent components.
図１Ａは、複数人の規格化元画像の色を平均化して得られる色平均顔について、見た目年齢の異なる顔画像を合成する方法の流れ図であり、図２この方法を実施する顔画像の合成装置１の一態様のブロック図である。 FIG. 1A is a flowchart of a method for synthesizing face images having different appearance ages with respect to a color average face obtained by averaging colors of a plurality of standardized original images, and FIG. 2 is a synthesis of face images for performing this method. 2 is a block diagram of one aspect of the device 1. FIG.
この顔画像の合成装置１は、元画像を取得する画像取得手段２と演算手段３を備えている。演算手段３は、高速コンピュータに、後述する標準顔の算出、画像の規格化、マスク処理、主成分分析、重回帰分析、画像の合成等に必要な演算機能と、画像や計算結果の保存機能等を組み込んだものであり、この演算手段３には、演算結果等を表示するディスプレイ４及びプリンタ５が接続されている。 The face image composition apparatus 1 includes an image acquisition unit 2 and an arithmetic unit 3 for acquiring an original image. The calculation means 3 is a high-speed computer that has a calculation function necessary for calculating a standard face, image normalization, mask processing, principal component analysis, multiple regression analysis, image synthesis, etc., and a function for storing images and calculation results. The display 4 and the printer 5 for displaying the calculation results are connected to the calculation means 3.
図１Ａに示す方法は、次の工程を有している。
Ａ．複数人の元画像を取得し、
Ｂ．その元画像の形状の特徴点が所定形状の顔型の特徴点に揃うように規格化した規格化元画像を作成し、
Ｃ．規格化元画像において、眉、目、口、毛髪及び背景をマスク処理することにより肌領域だけ抽出したマスク処理画像を作成し、
Ｄ．規格化元画像の色（より具体的には、マスク処理画像の肌領域のＲＧＢ画素値）について主成分分析し、
Ｅ．規格化元画像の見た目年齢のスコア値を取得し、
Ｆ．工程Ｄの主成分分析により得た規格化元画像のパラメータのうち、見た目年齢に寄与するものを特定し、
Ｇ．一方、複数人の規格化元画像のマスク処理画像からそれらの色平均顔を算出し、
Ｈ．色平均顔について、工程Ｆで求めた、見た目年齢に寄与するパラメータの重み係数を変化させることにより、色平均顔について、見た目年齢の異なる顔画像を得る。
The method shown in FIG. 1A includes the following steps.
A. Get the original images of multiple people
B. Create a standardized original image that is standardized so that the feature points of the shape of the original image are aligned with the face-shaped feature points of the predetermined shape,
C. In the standardized original image, create a masked image that only extracts the skin area by masking the eyebrows, eyes, mouth, hair and background,
D. The principal component analysis is performed on the color of the standardized original image (more specifically, the RGB pixel value of the skin area of the mask processing image),
E. Get the score value of the apparent age of the standardized original image,
F. Among the parameters of the standardized original image obtained by the principal component analysis of step D, identify those that contribute to the appearance age,
G. On the other hand, those color average faces are calculated from the masked images of the standardized original images of multiple people,
H. For the color average face, by changing the weighting coefficient of the parameter that contributes to the appearance age obtained in step F, a face image having a different appearance age is obtained for the color average face.
工程Ａで元画像の取得に使用する画像取得手段２としては、顔画像を撮るデジタルカメラ、顔写真から顔画像を取り込むイメージスキャナ、顔画像が記録されている任意の記録媒体から顔画像を取り込むドライブ、インターネット等の通信回線から顔画像を取り込む通信手段等をあげることができる。 Image acquisition means 2 used to acquire the original image in step A includes a digital camera that captures a face image, an image scanner that captures a face image from a face photograph, and a face image captured from any recording medium on which the face image is recorded. A communication means for capturing a face image from a communication line such as a drive or the Internet can be used.
ここで取り込む元画像は、素顔でも化粧顔でもよいが、一定照明条件下で取得することが好ましい。また、元画像の母集団の構成により、主成分分析によるパラメータの順位が大きく変わる場合があるため、元画像の母集団は、性別ごとにある年齢幅に属するように統制することが好ましく、例えば、女性で、18〜24歳、25〜39歳、又は40〜49歳等とすることが好ましい。また、最終的に取得する合成顔画像に付与したい顔型の特徴に応じて、二重瞼又は一重瞼等といった顔型の特徴ごとに元画像の母集団を統制してもよい。 The original image captured here may be a natural face or a makeup face, but is preferably acquired under a fixed illumination condition. In addition, since the order of parameters by principal component analysis may vary greatly depending on the composition of the population of the original image, it is preferable to control the population of the original image to belong to an age range for each gender, It is preferable that the female is 18-24 years old, 25-39 years old, 40-49 years old, or the like. Further, the population of the original image may be controlled for each facial feature such as a double eyelid or a single eyelid depending on the facial feature desired to be added to the finally obtained composite facial image.
演算手段３は、まず、画像取得手段２で取得した元画像を記憶する。 The calculation means 3 first stores the original image acquired by the image acquisition means 2.
工程Ｂにおいては、取り込んだ個々の元画像毎に、形状の特徴、例えば、毛髪のはえぎわ、ほほ、あご等の輪郭や眉、目、鼻、等の形状を特定し、それらの形状の特徴点をモーフィング処理により所定形状の顔型の特徴点に揃える。この規格化で使用する所定形状の顔型は、頭髪を含まない顔面のみのものとすることが好ましい。またこの顔型には、任意の顔面形状のものを使用することができ、実在の人物の顔型でも仮想の人物の顔型でもよく、複数人の元画像の形状のみを平均化した標準顔の顔型を使用してもよい。標準顔は特定人を全く表象しないため、特定人の画像であることを理由として利用が制限されることを解消できる。 In step B, for each captured original image, the shape characteristics, for example, the contours of hair, cheeks, chins, etc. and the shape of the eyebrows, eyes, nose, etc. are specified, and The feature points are aligned with face-shaped feature points of a predetermined shape by morphing processing. It is preferable that the face shape of a predetermined shape used in this normalization is only a face that does not include hair. In addition, this face type can be of any face shape, and may be a real person's face type or a virtual person's face type, and a standard face obtained by averaging only the shapes of the original images of a plurality of people. The face type may be used. Since the standard face does not represent the specific person at all, it is possible to eliminate the restriction of the use due to the fact that the standard face is an image of the specific person.
なお、ここで形状の特徴の特定や、モーフィング処理は、市販の顔画像合成システム（例えば、株式会社国際電気通信基礎研究所、futon）を用いて行うことができる。 Here, the feature of the shape and the morphing process can be performed using a commercially available face image synthesis system (for example, International Telecommunications Research Institute, Futon).
工程Ｃは、規格化元画像から、眉、目、口、毛髪及び背景の画素値を除去（マスク処理）したマスク処理画像を作成する工程である。眉、目、唇、毛髪及び背景はベースメイクの領域ではないことから、ベースメイクに強く関与するパラメータを得るため、本発明においては工程Ｃを行うことが好ましい。なお、このマスク処理は、規格化前の元画像に対して行っても良い。 Step C is a step of creating a mask processing image in which pixel values of eyebrows, eyes, mouth, hair, and background are removed (mask processing) from the standardized original image. Since eyebrows, eyes, lips, hair, and background are not base makeup regions, it is preferable to perform step C in the present invention in order to obtain parameters that are strongly involved in base makeup. This mask process may be performed on the original image before normalization.
このマスク処理により、例えば、図３に示すように、２５〜４０歳の複数人の素顔及び／又は化粧顔の元画像の形状を標準顔にモーフィング処理し、眉、目、唇、毛髪及び背景の画素値をマスク処理したマスク処理画像を得る。 By this masking process, for example, as shown in FIG. 3, the shape of the original image of the face of a plurality of people aged 25 to 40 and / or makeup face is morphed to a standard face, and the eyebrows, eyes, lips, hair and background A mask-processed image obtained by masking the pixel values is obtained.
工程Ｄでは、複数人の規格化元画像、好ましくはマスク処理した規格化元画像の色について主成分分析を行う。具体的には、複数人の規格化元画像の画像データ（Ｍ（横）×Ｎ（縦）画素）のＲＧＢ画素値を一列に並べたＬ（＝Ｍ×Ｎ）次元ベクトルを主成分分析し、元画像のサンプル数Ｑに等しいＱ次元の固有ベクトル（パラメータ）とスコア値（重み係数）を求める。 In step D, principal component analysis is performed on the colors of a plurality of standardized original images, preferably the masked standardized original images. Specifically, a principal component analysis is performed on an L (= M × N) dimensional vector in which RGB pixel values of image data (M (horizontal) × N (vertical) pixels) of standardized original images of a plurality of people are arranged in a line. Then, a Q-dimensional eigenvector (parameter) equal to the number of samples Q of the original image and a score value (weight coefficient) are obtained.
この主成分分析の手法は、M. Turk and A. Pentland, "Eigenfaces for recognition," Journal of Cognitive Neuroscience, vol. 3, no. 1, pp. 71-86, 1991に記載されている方法によることができ、例えばNETLIBの数値演算ライブラリCLAPACKを用いて行うことができる。 This principal component analysis method is based on the method described in M. Turk and A. Pentland, "Eigenfaces for recognition," Journal of Cognitive Neuroscience, vol. 3, no. 1, pp. 71-86, 1991. For example, it can be performed using the numerical calculation library CLAPACK of NETLIB.
次式(I)は、こうして得られた上位１７個のパラメータＸi(i＝1〜17)と重み係数ａi(i＝1〜17)で規格化元画像を表したものであり、図４は、上位１７次元のパラメータを画像として表現したものである。 The following equation (I) represents the standardized original image with the top 17 parameters X i (i = 1 to 17) and weighting factors a i (i = 1 to 17) obtained in this way. 4 represents the upper 17-dimensional parameters as an image.
一方、工程Ｅで、規格化元画像の見た目年齢のスコア値を取得する。スコア値としては、例えば、５〜２０人の美容専門家が見た目年齢を３〜１０歳刻みの年齢に評価したときの平均を算出する。 On the other hand, in step E, the score value of the apparent age of the standardized original image is acquired. As a score value, for example, an average is calculated when an appearance age of 5 to 20 beauty professionals is evaluated as an age in increments of 3 to 10 years.
工程Ｆでは、工程Ｄの主成分分析により得た規格化元画像のパラメータのうち、見た目年齢に寄与するものを特定する。この特定の手法としては、例えば、工程Ｅで得た見た目年齢のスコア値を目的変数とし、主成分分析により得られた規格化元画像のパラメータＸiの重み係数ａiを説明変数とする重回帰分析を行うにあたり、ステップワイズ法により見た目年齢に寄与するパラメータＸiの重み係数ａiを選択する。例えば、１６１人の２５〜４０歳の化粧顔を元画像として標準顔に規格化し、見た目年齢のスコア値を、１５人の美容専門家が１９歳以下、２０〜２４歳、２５〜２９歳、３０〜３４歳、３５〜３９歳、４０歳以上という６段階の基準で評価したときの平均値とすることにより、次の回帰式（II）を得た。 In step F, among the parameters of the standardized original image obtained by the principal component analysis in step D, those that contribute to the apparent age are specified. As this specific method, for example, the score value of the apparent age obtained in the step E is used as an objective variable, and the weighting factor a i of the parameter X i of the normalized original image obtained by the principal component analysis is used as an explanatory variable. In performing the regression analysis, the weight coefficient a i of the parameter X i that contributes to the apparent age is selected by the stepwise method. For example, 161 makeup faces of 25 to 40 years old are standardized as a standard face as an original image, and the score value of appearance age is determined by 15 beauty professionals under 19 years old, 20-24 years old, 25-29 years old, The following regression equation (II) was obtained by taking the average value when evaluated according to the six-stage criteria of 30-34 years old, 35-39 years old, and 40 years old or older.
式(II)からわかるように、見た目年齢に大きく寄与する重み係数は、パラメータＸ17、Ｘ14、Ｘ12、Ｘ11の重み係数ａ17、ａ14、ａ12、ａ11 であった。また、各パラメータの重み係数ａiの寄与を示す標準化偏回帰係数ｋ（絶対値の大きいものから選択される）はそれぞれｋ17は＋0.367、ｋ14は-0.319、ｋ12は-0.220、ｋ11は-0.217であった。 As can be seen from Formula (II), contributes significantly weighting factor to the eye ages were weighting factors a 17, a 14, a 12 , a 11 parameters X 17, X 14, X 12 , X 11. Also, k 17 respectively (as selected from the largest absolute value) standardized partial regression coefficient k indicating the contribution of the weighting factor a i for each parameter + 0.367, k 14 is -0.319, k 12 is -0.220, k 11 was -0.217.
なお、見た目年齢に寄与するパラメータの特定方法としては、上述の重回帰分析による方法の他、判別分析（小林敏和、大図正孝、大竹俊輔、赤松茂、"形状とテクスチャの特徴空間における線形判別関数を用いた顔イメージの生成", 日本顔学会誌４巻１号、pp.33-44, 2004年）等によってもよい。 In addition to the method of multiple regression analysis described above, discriminant analysis (Toshikazu Kobayashi, Masataka Otsu, Shunsuke Otake, Shigeru Akamatsu, “Linear discrimination in the feature space of shape and texture” "Generation of facial images using functions", Journal of the Japanese Facial Society Vol. 4, No. 1, pp. 33-44, 2004).
また、見た目年齢に寄与するパラメータとして特定するパラメータ数は、上述の４個に限られず、通常１〜２０個程度を特定することにより、効果的に見た目年齢を上下させることができる。また、このパラメータを特定する場合の条件を、例えば、重回帰分析の結果採用されるパラメータの累積寄与率が0.3以上というしきい値を設けることにより、パラメータを特定する機能を演算手段に持たせることができる。 Moreover, the number of parameters specified as a parameter contributing to the appearance age is not limited to the above-mentioned four, and it is possible to effectively increase or decrease the appearance age by specifying about 1 to 20 in general. In addition, as a condition for specifying this parameter, for example, by providing a threshold value that the cumulative contribution ratio of the parameter adopted as a result of the multiple regression analysis is 0.3 or more, the calculation means has a function of specifying the parameter. be able to.
工程Ｇでは、工程Ｃで作成した複数のマスク処理画像について、色を平均化した色平均顔を算出する。この色平均顔は、次の工程Ｈで顔画像の見た目年齢を変化させる際の基準となるものである。 In step G, a color average face obtained by averaging colors of the plurality of mask processed images created in step C is calculated. This color average face is a reference when changing the appearance age of the face image in the next step H.
工程Ｈでは、工程Ｇで算出した色平均顔について、工程Ｆで選択したパラメータの重み係数を変化させることにより、見た目年齢の異なる顔画像を作成する。より具体的には、例えば、前述の１６１人の化粧顔の色平均顔の場合、式(I)において、[規格化元画像の平均]を[色平均顔画像]とし、工程Ｆで選択したパラメータＸ17、Ｘ14、Ｘ12、Ｘ11 の重み係数ａ17、ａ14、ａ12、ａ11 を変化させることにより、色平均顔について見た目年齢を異ならせた顔画像を得る。 In Step H, face images having different appearance ages are created by changing the weighting coefficient of the parameter selected in Step F for the color average face calculated in Step G. More specifically, for example, in the case of the color average face of the 161 makeup faces described above, in [Equation (I)], [Average of standardized original image] is set to [Color average face image] and selected in step F. By changing the weight coefficients a 17 , a 14 , a 12 , and a 11 of the parameters X 17 , X 14 , X 12 , and X 11 , face images with different apparent ages are obtained for the color average face.
工程Ｆで得られた重回帰分析の標準化偏回帰係数ｋの符号により、それぞれの重み係数ａ17、ａ14、ａ12、ａ11 に対応するパラメータＸ17、Ｘ14、Ｘ12、Ｘ11 が、見た目年齢を高くするか低くするかという方向性がわかる。例えば、前述の式(II)によれば、見た目年齢を高くするためには、ａ17を正の方向に、ａ14、ａ12、ａ11を負の方向に変化させる。逆に、見た目年齢を低くするためには、ａ17を負の方向に、ａ14、ａ12、ａ11を正の方向に変化させる。 The sign of the standardized partial regression coefficient k of the multiple regression analysis obtained in step F, the parameter X 17, X 14, X 12 , X 11 corresponding to each of the weighting factors a 17, a 14, a 12 , a 11 , You can see the direction of increasing or decreasing the appearance age. For example, according to formula (II) described above, in order to increase the appearance age, a a 17 in the positive direction, a 14, a 12, changing the a 11 in the negative direction. Conversely, to lower the appearance age, a a 17 in the negative direction, changing the a 14, a 12, a 11 in the positive direction.
そこで、次式(IIIa)、式(IIIb)のように、見た目年齢を所定年齢分高く又は低くする成分画像を求め、この成分画像を[色平均顔画像]に重ね合わせる。 Therefore, as shown in the following formulas (IIIa) and (IIIb), a component image that increases or decreases the appearance age by a predetermined age is obtained, and this component image is superimposed on the [color average face image].
より具体的には、上述の式(IIIa)において、見た目年齢を５歳高くする成分画像は次式で得られる。 More specifically, in the above formula (IIIa), a component image that increases the apparent age by 5 years is obtained by the following formula.
また、上述の式(IIIb)において、見た目年齢を５歳低くする成分画像は次式で得られる。 In the above formula (IIIb), a component image that lowers the apparent age by 5 years is obtained by the following formula.
図５は、式(IIIa)、式(IIIb)に対応する成分画像と、そのパラメータの画像である。 FIG. 5 shows component images corresponding to the formulas (IIIa) and (IIIb) and their parameter images.
以上の色平均顔について見た目年齢を変化させる方法は、実在の顔や合成による顔を問わず任意の顔について、見た目年齢を変化させる場合に適用することができ、したがって、特定人が若く又は老けて見える画像を得るために使用することができる。 The above-mentioned method for changing the appearance age of the color average face can be applied to change the appearance age for any face regardless of the actual face or the face by synthesis, and therefore the specific person is young or old. Can be used to obtain a visible image.
その場合、図１Ｂに示すように、まず、工程Ａ’で、見た目年齢を異ならせようとする特定人の画像を取得し、工程Ｂ’でその特定人の規格化元画像を作成し、好ましくは工程Ｃ’で、規格化元画像として、工程Ｂ’で得た規格化元画像のマスク処理画像を作成する。 In that case, as shown in FIG. 1B, first, in step A ′, an image of a specific person who wants to change the appearance age is acquired, and in step B ′, a standardized original image of the specific person is created, In step C ′, a masked image of the standardized original image obtained in step B ′ is created as a standardized original image.
一方、前述の色平均顔について、見た目年齢を変化させる場合と同様にして、複数人の元画像から、見た目年齢に寄与するパラメータを特定する（工程Ａ〜工程Ｆ）。そして、工程Ｈ’において、次式Ｉ'で表される特定人のマスク処理画像について、工程Ｆで特定したパラメータの重み係数を変化させ、見た目年齢が高い又は低い規格化顔画像を形成する。 On the other hand, for the color average face, parameters that contribute to the appearance age are specified from the original images of a plurality of people in the same manner as when the appearance age is changed (steps A to F). In step H ′, the weighted coefficient of the parameter specified in step F is changed for the mask processing image of the specific person represented by the following formula I ′ to form a standardized face image with a high or low visual age.
あるいは、見た目年齢を高くする成分画像（式(IIIa)）、又は低くする成分画像（式(IIIb)）を求め、図６のように、特定人の規格化元画像に式(IIIa)又は式(IIIb)の成分画像を加え、見た目年齢が高い又は低い規格化顔画像を形成する。 Alternatively, a component image (formula (IIIa)) for increasing the appearance age or a component image (formula (IIIb)) for increasing the appearance age is obtained, and the formula (IIIa) or formula is added to the standardized original image of a specific person as shown in FIG. The component image of (IIIb) is added to form a standardized face image with a high or low visual age.
次に、工程Ｉで、見た目年齢を変化させた規格化顔画像の特徴点を特定人の元画像の特徴点に戻す（即ち、Ａ’からＢ‘で行った工程、又はＡ’からＣ‘で行った工程の逆の操作を行う）。こうして合成された顔画像は、顔の輪郭や、目、鼻、口の位置や形状といった顔の基本設計は特定人の元画像と共通だが、見た目年齢が異なったものとなる。 Next, in step I, the feature points of the standardized face image whose appearance age is changed are returned to the feature points of the original image of the specific person (that is, the steps performed from A ′ to B ′, or A ′ to C ′). The reverse of the process performed in step 1). The face image synthesized in this way has the same basic design as the original image of a specific person, such as the outline of the face and the position and shape of the eyes, nose, and mouth, but the appearance age is different.
元画像と、本発明により得られる顔画像との見た目年齢の差は、化粧を変えることにより得られるものと解釈することができるため、本発明の方法は、化粧方法、化粧品素材、化粧料等の研究、開発や、化粧方法のアドバイス等の分野で有用となる。 Since the difference in appearance age between the original image and the face image obtained by the present invention can be interpreted as being obtained by changing makeup, the method of the present invention is a cosmetic method, cosmetic material, cosmetics, etc. It is useful in fields such as research, development, and advice on makeup methods.
１ 顔画像の合成装置
２ 画像取得手段
３ 演算手段
４ ディスプレイ
５ プリンタ
DESCRIPTION OF SYMBOLS 1 Face image composition apparatus 2 Image acquisition means 3 Calculation means 4 Display 5 Printer
Claims (10)
複数人の顔画像（以下、元画像という）を取得し、
各元画像から、その形状の特徴点が所定形状の顔型の特徴点に揃うように規格化した画像（以下、規格化元画像という）を作成し、
複数人の規格化元画像の色について主成分分析を行い、
主成分分析により得られる規格化元画像の固有ベクトル（以下、パラメータという）中、見た目年齢に寄与するパラメータを特定し、
一方、任意の顔の規格化元画像を作成し、任意の顔の規格化元画像において前記主成分分析で得られる固有ベクトル空間における主成分スコア（以下、パラメータの重み係数という）を変化させることにより見た目年齢の異なる規格化顔画像を作成する顔画像の合成方法。 A method of compositing facial images with different appearance ages,
Acquire face images of multiple people (hereinafter referred to as original images)
From each original image, create a standardized image (hereinafter referred to as the standardized original image) so that the feature points of the shape are aligned with the face-shaped feature points of the predetermined shape
Perform principal component analysis on the colors of the standardized original images of multiple people,
Among the eigenvectors (hereinafter referred to as parameters) of the standardized original image obtained by principal component analysis, identify parameters that contribute to visual age,
On the other hand, by creating a standardized original image of an arbitrary face and changing a principal component score (hereinafter referred to as a parameter weighting factor) in the eigenvector space obtained by the principal component analysis in the standardized original image of the arbitrary face A face image synthesis method for creating standardized face images having different appearance ages.
前記主成分分析を複数人のマスク処理画像の色について行う請求項１又は２記載の顔画像の合成方法。 As a standardized original image, after normalizing the original image into a face shape of a predetermined shape, an image (hereinafter referred to as a masked image) in which the eyebrows, eyes, mouth, hair and background are masked is created,
The face image synthesis method according to claim 1, wherein the principal component analysis is performed for colors of a plurality of masked images.
該演算手段が、
元画像から、その形状の特徴点が所定形状の顔型の特徴点に揃うように規格化した画像（以下、規格化元画像という）を作成する機能、
複数人の規格化元画像の色について主成分分析を行う機能、及び
主成分分析による規格化元画像の固有ベクトル（以下、パラメータという）中、見た目年齢に寄与するパラメータを特定する機能、
任意の顔の規格化元画像において、前記主成分分析で得られる固有ベクトル空間における主成分スコア（以下、パラメータの重み係数という）を変化させることにより、見た目年齢の異なる規格化顔画像を作成する機能を備えている顔画像の合成装置。 A face image composition device comprising: an image acquisition means for obtaining a face image (hereinafter referred to as an original image); and a calculation means for creating a composite face image having a different apparent age from the original image,
The computing means is
A function that creates an image (hereinafter referred to as a normalized original image) that is standardized so that the feature points of the shape are aligned with the face-shaped feature points of a predetermined shape
A function for performing principal component analysis on the colors of the standardized original images of a plurality of people, and a function for specifying a parameter that contributes to an apparent age among eigenvectors (hereinafter referred to as parameters) of the standardized original image by principal component analysis,
A function for creating a standardized face image having a different appearance age by changing a principal component score (hereinafter referred to as a parameter weighting factor) in an eigenvector space obtained by the principal component analysis in a standardized original image of an arbitrary face An apparatus for synthesizing a face image.
を備えている請求項７記載の顔画像の合成装置。 Further, a function of forming a composite face image having a different appearance age for the arbitrary face by returning the feature points of the standardized face image having a different appearance age to the feature points of the original image of the arbitrary face,
The face image composition device according to claim 7.
前記所定形状の顔型として、標準顔を使用する請求項７又は８記載の顔画像の合成装置。 The computing means has a function of calculating a standard face obtained by averaging the shapes of the original images of a plurality of people,
The face image composition device according to claim 7 or 8, wherein a standard face is used as the face shape of the predetermined shape.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2008148648A JP4893968B2 (en) | 2008-06-05 | 2008-06-05 | How to compose face images |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| JP2008148648A JP4893968B2 (en) | 2008-06-05 | 2008-06-05 | How to compose face images |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| JP2009294958A true JP2009294958A (en) | 2009-12-17 |
| JP4893968B2 JP4893968B2 (en) | 2012-03-07 |
Family
ID=41543081
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| JP2008148648A Active JP4893968B2 (en) | 2008-06-05 | 2008-06-05 | How to compose face images |
Country Status (1)
| Country | Link |
|---|---|
| JP (1) | JP4893968B2 (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2012008617A (en) * | 2010-06-22 | 2012-01-12 | Kao Corp | Face image evaluation method, face evaluation method and image processing device |
| WO2015125759A1 (en) * | 2014-02-24 | 2015-08-27 | 花王株式会社 | Aging analysis method and aging analysis device |
| WO2018131105A1 (en) * | 2017-01-12 | 2018-07-19 | Nec Corporation | Information processing device, information processing method and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2005293231A (en) * | 2004-03-31 | 2005-10-20 | Kao Corp | Simulation image forming method for skin |
| JP2007128171A (en) * | 2005-11-01 | 2007-05-24 | Advanced Telecommunication Research Institute International | Face image synthesizer, face image synthesizing method and face image synthesizing program |
| WO2009090948A1 (en) * | 2008-01-17 | 2009-07-23 | Shiseido Company, Ltd. | Skin color evaluation method, skin color evaluation apparatus, skin color evaluation program, and recording medium with the program recorded thereon |
-
2008
- 2008-06-05 JP JP2008148648A patent/JP4893968B2/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2005293231A (en) * | 2004-03-31 | 2005-10-20 | Kao Corp | Simulation image forming method for skin |
| JP2007128171A (en) * | 2005-11-01 | 2007-05-24 | Advanced Telecommunication Research Institute International | Face image synthesizer, face image synthesizing method and face image synthesizing program |
| WO2009090948A1 (en) * | 2008-01-17 | 2009-07-23 | Shiseido Company, Ltd. | Skin color evaluation method, skin color evaluation apparatus, skin color evaluation program, and recording medium with the program recorded thereon |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2012008617A (en) * | 2010-06-22 | 2012-01-12 | Kao Corp | Face image evaluation method, face evaluation method and image processing device |
| WO2015125759A1 (en) * | 2014-02-24 | 2015-08-27 | 花王株式会社 | Aging analysis method and aging analysis device |
| JP2015172935A (en) * | 2014-02-24 | 2015-10-01 | 花王株式会社 | Aging analysis method and aging analysis device |
| CN106030659A (en) * | 2014-02-24 | 2016-10-12 | 花王株式会社 | Aging analysis method and aging analysis device |
| CN106030659B (en) * | 2014-02-24 | 2019-01-22 | 花王株式会社 | Increase age analysis method and increases age analytical equipment |
| TWI716344B (en) * | 2014-02-24 | 2021-01-21 | 日商花王股份有限公司 | Aging analyzing method, aging care counselling method using aging analyzing method, aging analyzing device and computer readable recording medium |
| WO2018131105A1 (en) * | 2017-01-12 | 2018-07-19 | Nec Corporation | Information processing device, information processing method and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| JP4893968B2 (en) | 2012-03-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Laurentini et al. | Computer analysis of face beauty: A survey | |
| EP2174296B1 (en) | Method and apparatus for realistic simulation of wrinkle aging and de-aging | |
| US20100189357A1 (en) | Method and device for the virtual simulation of a sequence of video images | |
| WO2013005447A1 (en) | Face impression analysis method, cosmetic counseling method, and face image generation method | |
| JP6207210B2 (en) | Information processing apparatus and method | |
| WO2007063878A1 (en) | Face classifying method, face classifying device, classification map, face classifying program, recording medium where this program is recorded | |
| JP2001346627A (en) | Make-up advice system | |
| JP5035524B2 (en) | Facial image composition method and composition apparatus | |
| JP2007128171A (en) | Face image synthesizer, face image synthesizing method and face image synthesizing program | |
| JP3920747B2 (en) | Image processing device | |
| JP5651385B2 (en) | Face evaluation method | |
| JP2000194835A (en) | Makeup advice system | |
| CN113344837B (en) | Face image processing method and device, computer readable storage medium and terminal | |
| JP4893968B2 (en) | How to compose face images | |
| JP5095182B2 (en) | Face classification device, face classification program, and recording medium on which the program is recorded | |
| KR20020014844A (en) | Three dimensional face modeling method | |
| KR20020007744A (en) | Modeling method of future children and future facial shapes | |
| Varshovi | Facial makeup detection using HSV color space and texture analysis | |
| Mena-Chalco et al. | 3D human face reconstruction using principal components spaces | |
| JP5650012B2 (en) | Facial image processing method, beauty counseling method, and facial image processing apparatus | |
| JPH1021290A (en) | Makeup simulating method | |
| KR20020007735A (en) | Three dimensional face modeling method and facial photo sticker system for predicting the future children and the future facial shapes | |
| Leta et al. | A study of the facial aging-a multidisciplinary approach | |
| Nomura et al. | Canonical correlation analysis for long-term changes of facial images based on the frequency of UV protection, physical and psychological features | |
| Jayasinghe et al. | Matching facial images using age related morphing changes |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| A621 | Written request for application examination |
Free format text: JAPANESE INTERMEDIATE CODE: A621 Effective date: 20110301 |
|
| A977 | Report on retrieval |
Free format text: JAPANESE INTERMEDIATE CODE: A971007 Effective date: 20111121 |
|
| TRDD | Decision of grant or rejection written | ||
| A01 | Written decision to grant a patent or to grant a registration (utility model) |
Free format text: JAPANESE INTERMEDIATE CODE: A01 Effective date: 20111124 |
|
| A01 | Written decision to grant a patent or to grant a registration (utility model) |
Free format text: JAPANESE INTERMEDIATE CODE: A01 |
|
| A61 | First payment of annual fees (during grant procedure) |
Free format text: JAPANESE INTERMEDIATE CODE: A61 Effective date: 20111207 |
|
| R151 | Written notification of patent or utility model registration |
Ref document number: 4893968 Country of ref document: JP Free format text: JAPANESE INTERMEDIATE CODE: R151 |
|
| FPAY | Renewal fee payment (event date is renewal date of database) |
Free format text: PAYMENT UNTIL: 20150106 Year of fee payment: 3 |
|
| R250 | Receipt of annual fees |
Free format text: JAPANESE INTERMEDIATE CODE: R250 |
|
| R250 | Receipt of annual fees |
Free format text: JAPANESE INTERMEDIATE CODE: R250 |
|
| R250 | Receipt of annual fees |
Free format text: JAPANESE INTERMEDIATE CODE: R250 |
|
| R250 | Receipt of annual fees |
Free format text: JAPANESE INTERMEDIATE CODE: R250 |
|
| R250 | Receipt of annual fees |
Free format text: JAPANESE INTERMEDIATE CODE: R250 |
|
| R250 | Receipt of annual fees |
Free format text: JAPANESE INTERMEDIATE CODE: R250 |
|
| R250 | Receipt of annual fees |
Free format text: JAPANESE INTERMEDIATE CODE: R250 |
|
| R250 | Receipt of annual fees |
Free format text: JAPANESE INTERMEDIATE CODE: R250 |
|
| R250 | Receipt of annual fees |
Free format text: JAPANESE INTERMEDIATE CODE: R250 |
|
| R250 | Receipt of annual fees |
Free format text: JAPANESE INTERMEDIATE CODE: R250 |