- 最后登录
- 2017-9-18
- 注册时间
- 2011-1-12
- 阅读权限
- 90
- 积分
- 12276
  
- 纳金币
- 5568
- 精华
- 0
|
1 Introduction
Most conventional example-based texture transfer methods have
taken a pixel- or patch-based approach to synthesizing the output
results [Hertzmann et al. 2001; Lee et al. 2010; Wang et al. 2004].
A patch of the source example, which must be big enough to provide
useful information, can express a localize “mixture of vision”
by itself only. The patch-based approach might suit an Impressionist
painterly style transfer, but most texture transfers use only the
luminance values during synthesis and keep all other color components
of the target image unchanged. Color expression is one of the
most important factors for Impressionist paintings, so ideally the
separation of luminance values from the other color components of
a source example should be avoided. In this paper, we propose a
patch-based texture transfer using both the luminance and the color
values of a source image. Experimental results showed that the proposed
method can produce an output result visually quite similar to
the Impressionist painting source.
2 Proposed method
The proposed method requires two input images: a source example
and a target image, as shown in Fig. 1. First, the target is represented
as a set of blocks divided recursively using an edge-based
distance transform [Hara et al. 2005]. Second, a user chooses one
or more initial seed patches from the source and manually assigns
them to various places on the target. Third, the minimization of
Eq. (1) automatically determines which source patches to assign
to which neighbors. The seeds thus grow and expand on the target.
For the neighbors of the expanded seeds, Eq. (1) is iterated
to assign additional source patches, and the resulting image is then synthesized. |
|