Processing

Please wait...

Settings

Settings

Goto Application

1. WO2020063527 - HUMAN HAIRSTYLE GENERATION METHOD BASED ON MULTI-FEATURE RETRIEVAL AND DEFORMATION

Publication Number WO/2020/063527
Publication Date 02.04.2020
International Application No. PCT/CN2019/107263
International Filing Date 23.09.2019
IPC
G06F 16/53 2019.01
GPHYSICS
06COMPUTING; CALCULATING OR COUNTING
FELECTRIC DIGITAL DATA PROCESSING
16Information retrieval; Database structures therefor; File system structures therefor
50of still image data
53Querying
CPC
G06K 9/00268
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
KRECOGNITION OF DATA; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
9Methods or arrangements for reading or recognising printed or written characters or for recognising patterns, e.g. fingerprints
00221Acquiring or recognising human faces, facial parts, facial sketches, facial expressions
00268Feature extraction; Face representation
G06K 9/6212
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
KRECOGNITION OF DATA; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
9Methods or arrangements for reading or recognising printed or written characters or for recognising patterns, e.g. fingerprints
62Methods or arrangements for recognition using electronic means
6201Matching; Proximity measures
6212Comparing statistics of pixel or of feature values, e.g. histogram matching
G06K 9/6215
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
KRECOGNITION OF DATA; PRESENTATION OF DATA; RECORD CARRIERS; HANDLING RECORD CARRIERS
9Methods or arrangements for reading or recognising printed or written characters or for recognising patterns, e.g. fingerprints
62Methods or arrangements for recognition using electronic means
6201Matching; Proximity measures
6215Proximity measures, i.e. similarity or distance measures
G06T 2207/30201
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
2207Indexing scheme for image analysis or image enhancement
30Subject of image; Context of image processing
30196Human being; Person
30201Face
G06T 7/337
GPHYSICS
06COMPUTING; CALCULATING; COUNTING
TIMAGE DATA PROCESSING OR GENERATION, IN GENERAL
7Image analysis
30Determination of transform parameters for the alignment of images, i.e. image registration
33using feature-based methods
337involving reference images or patches
Applicants
  • 叠境数字科技(上海)有限公司 PLEX-VR DIGITAL TECHNOLOGY (SHANGHAI) CO., LTD. [CN]/[CN]
Inventors
  • 蒋琪雷 JIANG, Qilei
  • 马原曦 MA, Yuanxi
  • 张迎梁 ZHANG, Yingliang
Agents
  • 上海智晟知识产权代理事务所(特殊普通合伙) SHANGHAI ZHI-SHENG INTELLECTUAL PROPERTY OFFICE
Priority Data
201811165895.930.09.2018CN
Publication Language Chinese (ZH)
Filing Language Chinese (ZH)
Designated States
Title
(EN) HUMAN HAIRSTYLE GENERATION METHOD BASED ON MULTI-FEATURE RETRIEVAL AND DEFORMATION
(FR) PROCÉDÉ DE GÉNÉRATION DE COIFFURES SUR LA BASE D’UNE RÉCUPÉRATION ET DÉFORMATION DE MULTIPLES CARACTÉRISTIQUES
(ZH) 基于多特征检索和形变的人体发型生成方法
Abstract
(EN)
A human hairstyle generation method based on multi-feature retrieval and deformation, comprising: obtaining a hairstyle mask; recognizing feature points of a face and matching same with feature points in a hairstyle database; aligning an image with a standard face to obtain a corresponding hair area; calculating a minkowski distance of hair masks of all frontal faces in the hair region and hairstyle database; providing corresponding weight after sorting from small to large; training a deep learning network to detect the hairstyles of hair basic blocks at different scales; and taking out the most similar picture of hair. The present invention utilizes a single frontal face photo to retrieve a three-dimensional hair model which is most similar to the photo in a large three-dimensional hairstyle database by means of a retrieval database, thereby avoiding manual modeling, improving efficiency and ensuring high fidelity.
(FR)
Un procédé de génération de coiffures sur la base d’une récupération et d’une déformation de multiples caractéristiques comprend : l’obtention d’un masque de coiffure ; la reconnaissance de points caractéristiques d’un visage et l’appariement de ceux-ci à des points caractéristiques dans une base de données de coiffures ; l’alignement d’une image avec un visage standard pour obtenir une zone de chevelure correspondante ; le calcul d’une distance de Minkowski de masques de chevelure de tous les visages vus de face dans la région de chevelure et dans la base de données de coiffures ; la fourniture d’un poids correspondant après le tri de petit à grand ; la formation d’un réseau d’apprentissage profond pour détecter les coiffures de blocs de base de chevelure à différentes échelles ; et le prélèvement de l’image de chevelure la plus similaire. La présente invention utilise une seule photographie d’un visage vu de face pour récupérer un modèle de chevelure tridimensionnelle qui est le plus similaire à la photographie dans une grande base de données de coiffures tridimensionnelles au moyen d’une base de données de récupération, ce qui évite une modélisation manuelle, améliore l’efficacité et assure une grande fidélité.
(ZH)
基于多特征检索和形变的人体发型生成方法,包括:获取发型掩膜;识别人脸的特征点,和发型数据库中的匹配;将图像和标准人脸对齐,得到相应的头发区域;计算所述头发区域和发型数据库中所有正面人脸的头发掩膜的闵氏距离;并在从小到大排序后赋予相应的权重;训练深度学习网络,检测不同尺度下头发基础块的发型;取出最相似的头发图片。本发明利用单张正面人脸照片,通过检索数据库在大型的三维发型数据库检索与照片最相似的三维头发模型,避免手动建模,提高了效率,保证了较高的逼真度。
Also published as
Latest bibliographic data on file with the International Bureau