Portrait Editing on Face Fusion Using LP
Fan Huang, Dongming Zhou
*
, Rencan Nie, Guanghui Cai, Kangjian He, Quan Wang
Information College
Yunnan University
Kunming 650500, China
e-mail: zhoudm@ynu.edu.cn
Abstract—Aiming at the current medical cosmetology, late into
the film and television both photo editing and predicting the
baby looks specific application in the field of entertainment,
this paper proposes a face fusion technology based on
Laplacian pyramid (LP). Firstly, Harris corner detection
technology is introduced in the technology whether source face
image and the target face image features at the same level, and
the features are extracted. Then according to the requirements
of reality, the needs of the face portrait are generated by
reconstructing LP algorithm. Finally, according to the fusion
effect, the shadow compensation is executed. Experimental
results show that this technique is efficient and can adapt to
various face editing conditions and requirements, and the
fusion effects meet the practical requirements.
Keywords-laplacian pyramid; face fusion; harris corner
detection; shadow compensation Introduction
I. INTRODUCTION
Face is an important communication channel. It is a
challenging problem that the fusion of faces from different
people creates a new image in image processing, medical
cosmetology and entertainment field. Generally speaking, it
is required to replace or merge with another individual's
specific features to meet the needs of the demanders. If we
use traditional image editing technology, although the
synthetic results are eligible, the operation is tedious and
time-consuming. Now according to the relationship of source
image and target image, this is roughly divided into two
categories, as shown in Figure 1.
Target
face
Source
face
Cartoonlization
Editing
Figure 1. Sketch map of face editing.
The first category is the direct replacement of facial
features in the details. The replacement is mainly to
automatically seamless clone target image to source image in
the specific region [1]. Gao Yan proposed a local organ
replacement algorithm, using the Laplacian operator as a
constrain, to mix the target features with the original features
replacement, according to the correspondence of facial
feature points, a global and local facial detail migration
model is proposed by Song Mingli [2]. Liu Jinyun proposed
a novel expressions algorithm, which is assigned different
transmission models for different pieces of target face
identity information and subsequently introduce interaction
[3]. Agarwala combined the method of graph-cut
optimization, gradient domain and Possion, users only need
to draw a number of lines in the front area and the
background area to complete the interactive image editing
[4]. There are still lots of similar methods in the combination
[5]. These methods are highly efficient, but require
occasional user intervention. It will also have distortion of
the situation cannot be achieved fully automated, robust bad.
The second category is the feature matching of target
photos to find the relatively prominent characteristics. Using
the method of the combination of model extract features with
characteristics of the line pair, to realize automatic
deformation on the outstanding characteristic and generate
cartoon portraits or CG movie characters. In the field of non
realistic image processing, Brennan [6] proposed a set of
systematic from interactive methods to produce exaggerated
images when facial features need to preserve and exaggerate.
Koshimizu [7] proposed a PICASSO system, which is a
typical KANSEI vision system for facial caricature, but it is
blunt tone. Yang [8] investigated a user-friendly caricature
generation system based on exaggerating the difference from
the mean face, which allows a user to freely control each
generation step and design his or her own unique caricature
portrait to simplify the facial contour, in the database to
select a similar facial cartoon image combination, and
generate a cartoon portrait, but need to define a number of
categories in advance. Liang [9] presented a system that
automatically generates caricatures from input face images.
The exaggeration of a caricature is accomplished by a
prototype-based method that captures the artist's
understanding of what are distinctive features of a face and
the exaggeration style. These methods are now mature and
perfect, so the algorithm proposed in this paper is based on
the first class of human face image editing.
This paper first uses Harris corner detection to detect and
match the feature points between the target image and the
source image of facial features, and then the Laplacian
pyramid fusion is executed between target image and source
image. According the results of fusion image, the last step is
1328
2017 9th IEEE International Conference on Communication Software and Networks
978-1-5090-3820-6/17/$31.00 ©2017 IEEE