- 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
Color transfer is a practical image editing technology which is useful
in various applications. An ideal color transfer algorithm should
keep the scene in the source image and apply the color styles of
the reference image. All the dominant color styles of the reference
image should be presented in the result especially when there are
similar contents in the source and reference images.
Reinhard et al. [2001] present a global color transfer algorithm
which manipulates the pixel value of source image to match the
global mean and standard deviation of the reference image in a
uncorrelated color space l. Unnatural results may be produced
when the color distributions of source and reference images are very
different. Dong et al. [2010] extract the same number of dominant
color descriptor (DCD) from the input images and find an optimal
one-to-one mapping between the dominant color sets. All the obvious
color styles of the reference image can be presented in the
result if there are enough corresponding dominant color patterns
in the source. However, as shown in Figure 1, the normal local
color transfer algorithms will fail to generate a satisfied result if
the number of source dominant colors is much less than the reference
one, especially when the color styles are transferred between
similar contents. Moreover, the spatial distribution of the reference
color pattern should also be considered during the transfer process.
We propose a robust color transfer framework to address the above
issues. Our method can establish a soft connections between the
local color statistics of the source and reference images. All the
obvious color features can be presented in the result image, as well
as the spatial distribution of the reference color pattern. |
|