120 IEEE TRANSACTIONS ON IMAGE PROCESSING, VOL. 24, NO. 1, JANUARY 2015
Weighted Guided Image Filtering
Zhengguo Li, Senior Member, IEEE, Jinghong Zheng, Member, IEEE, Zijian Zhu, Member, IEEE,
Wei Yao, Member, IEEE, and Shiqian Wu, Senior Member, IEEE
Abstract—It is known that local filtering-based edge-
preserving smoothing techniques suffer from halo artifacts.
In this paper, a weighted guided image filter (WGIF) is intro-
duced by incorporating an edge-aware weighting into an existing
guided image filter (GIF) to address the problem. The WGIF
inherits advantages of both global and local smoothing filters in
the sense that: 1) the complexity of the WGIF is O(N) for an
image with N pixels, which is same as the GIF and 2) the WGIF
can avoid halo artifacts like the existing global smoothing filters.
The WGIF is applied for single image detail enhancement, single
image haze removal, and fusion of differently exposed images.
Experimental results show that the resultant algorithms produce
images with better visual quality and at the same time halo
artifacts can be reduced/avoided from appearing in the final
images with negligible increment on running times.
Index Terms—Edge-preserving smoothing, weighted guided
image filter, edge-aware weighting, detail enhancement, haze
removal, exposure fusion.
I. INTRODUCTION
M
ANY applications in the fields of computational
photography and image processing require smoothing
techniques that can preserve edge well. Typical examples
include image de-noising [1], [2], fusion of differently exposed
images [3], tone mapping of high dynamic range (HDR)
images [4], detail enhancement via multi-lighting images [5],
texture transfer from a source image to a destination image [6],
single image haze removal [7], and etc. The smoothing process
usually decomposes an image to be filtered into two layers:
a base layer formed by homogeneous regions with sharp edges
and a detail layer which can be either noise, e.g., a random
pattern with zero mean, or texture, such as a repeated pattern
with regular structure.
There are two types of edge-preserving image smoothing
techniques. One type is global optimization based filters as
in [1], [2], [4], and [8]. The optimized performance criterion
consists of a data term and a regularization term. The data
Manuscript received April 28, 2014; revised September 8, 2014 and
October 20, 2014; accepted November 10, 2014. Date of publication Novem-
ber 14, 2014; date of current version December 8, 2014. The associate editor
coordinating the review of this manuscript and approving it for publication
was Dr. Xin Li.
Z. Li, J. Zheng, Z. Zhu, and W. Yao are with the Department of
Signal Processing, Institute for Infocomm Research, Singapore 138632
(e-mail: ezgli@i2r.a-star.edu.sg; jzheng@i2r.a-star.edu.sg; zhuzj@i2r.a-star.
edu.sg; wyao@i2r.a-star.edu.sg).
S. Wu is with the College of Machinery and Automation, Wuhan
University of Science and Technology, Wuhan 430081, China (e-mail:
shiqian.wu@wust.edu.cn).
Color versions of one or more of the figures in this paper are available
online at http://ieeexplore.ieee.org.
Digital Object Identifier 10.1109/TIP.2014.2371234
term measures fidelity of reconstructed image with respect
to the image to be filtered while the regularization term
provides the smoothness level of the reconstructed image.
Even though the global optimization based filters often yield
excellent quality, they have high computational cost. The other
type is local filters such as bilateral filter (BF) [9], its
extension in gradient domain [10], trilateral filter [11], and
their accelerated versions [5], [12], [13] as well as guided
image filter (GIF) [14]. Compared with the global optimization
based filters, the local filters are generally simpler. However,
the local filters cannot preserve sharp edges like the global
optimization based filters [4], [14]. As such, halo artifacts are
usually produced by the local filters when they are adopted
to smooth edges [14]. It was mentioned in [14] that the local
filters such as the BF/GIF would concentrate the blurring near
these edges and introduce halos while the global optimization
based filters such as the weighted least squares (WLS) filter
in [4] would distribute such blurring globally. It is worth noting
that the Lagrangian factor in the WLS filter [4] is content
adaptive whether the Lagrangian factor in the GIF and both
spatial similarity parameter and range similarity parameter in
the BF [9] are fixed. This could be another major reason that
the BF/GIF produces halo artifacts. It is worth noting that the
reason was also noticed in [15] and [16]. The range similarity
parameter of the BF in [15] is adaptive to the content of
the image to be filtered while both the spatial similarity and
the range similarity parameters of the BF in [16] are adaptive
to the content of the image to be filtered. Unfortunately,
as pointed out in [15], adaptation of the parameters will
destroy the 3D convolution form, and the adaptive BF (ABF)
cannot be accelerated via the approach in [13]. It is thus
desired to design a new local filter which is as fast as the GIF
in [14] and preserves edges as well as the WLS filter in [4].
In this paper, an edge-aware weighting is introduced
and incorporated into the GIF [14] to form a
weighted GIF (WGIF). In human visual perception, edges
provide an effective and expressive stimulation that is vital for
neural interpretation of a scene [17]. Larger weights are thus
assigned to pixels at edges than pixels in flat areas. There are
many methods to compute the edge-ware weighting. Local
variance in 3 × 3 window of a pixel in a guidance image is
applied to compute the edge-aware weighting. The weighting
can be easily computed via the box filter in [14] for all
pixels in the guidance image. The local variance of a pixel is
normalized by the local variances of all pixels in the guidance
image. The normalized weighting is then adopted to design
the WGIF. Due to the proposed weighting, the WGIF can
preserve sharp edges like the global filters [1], [2], [4], [8].
1057-7149 © 2014 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
下载后可阅读完整内容，剩余9页未读，立即下载
Shipley_Xie
- 粉丝: 0
- 资源: 9
我的内容管理
展开
最新资源
- WebLogic集群配置与管理实战指南
- AIX5.3上安装Weblogic 9.2详细步骤
- 面向对象编程模拟试题详解与解析
- Flex+FMS2.0中文教程：开发流媒体应用的实践指南
- PID调节深入解析：从入门到精通
- 数字水印技术：保护版权的新防线
- 8位数码管显示24小时制数字电子钟程序设计
- Mhdd免费版详细使用教程：硬盘检测与坏道屏蔽
- 操作系统期末复习指南：进程、线程与系统调用详解
- Cognos8性能优化指南：软件参数与报表设计调优
- Cognos8开发入门：从Transformer到ReportStudio
- Cisco 6509交换机配置全面指南
- C#入门：XML基础教程与实例解析
- Matlab振动分析详解：从单自由度到6自由度模型
- Eclipse JDT中的ASTParser详解与核心类介绍
- Java程序员必备资源网站大全
资源上传下载、课程学习等过程中有任何疑问或建议，欢迎提出宝贵意见哦~我们会及时处理！
点击此处反馈
