Cluster-Based Cross-view Filtering for Compressed
Multi-view Depth Maps
Zhen Liu
#1
, Qiong Liu
>2
, You Yang
>2
, Yuchi Liu
∗3
, Gangyi Jiang
#1
, Mei Yu
#1
#1
Faculty of Information Science and Engineering,
Ningbo University, Ningbo, China
>2
School of Electronic Information and Communications,
Huazhong University of Science and Technology, Hubei, China
∗3
Early Warning Institute, Hubei, China
1
liuzhenuse@126.com
2
q.liu@hust.edu.cn
3
yangyou@hust.edu.cn
4
bill liew@163.com
5
jianggangyi@126.com
6
yumei2@126.com
Abstract—In the field of multi-view video coding, multi-view
plus depth video is an important data format, but it always
suffers from quantization errors, which result in obvious arti-
facts in consequent virtual view rendering. In this paper, we
propose a cluster-based cross-view filtering (CBF) scheme for
the enhancement of compressed depth maps. In this scheme,
reconstructed depth information are mapped from cross-view,
and this information is benefit to the proposed filter. Then in
filtering one viewpoint depth map with candidate information
that are selected from non-locally current and neighboring
viewpoints. Specifically, in our scheme, candidates are clustered
in 3D super-pixel wise rather than block wise due to cross-
relationship among pixels in depth maps. The experimental
results show that 2.0074 dB average gain can be obtained by
our scheme, which suggests that the scheme outperforms than
state-of-the-art and classical filters in filtering the reconstructed
depth maps.
Index Terms—depth image filter, depth map reconstructed,
multi-view video coding, cluster-based filter,coding distortion
I. I
Arbitrary virtual views synthesis is an amazing functionality
of free-view video (FVV) system [1]. In this case, multi-
view color videos and depth maps are both required, and then
becomes an important data format in multi-view video coding.
The 3D video extension of High Efficiency Video Coding
(HEVC) [2] enables the above representations and provides
a number of new coding tools for depth video. These tools
are benefit for high compression ratio, but coding distortions
is inevitable on the reconstructed depth maps [3]. Depth maps
are characterized by large homogeneous areas and sharp edges.
Actually, these features are sensitive to compression tools.
Fig. 1 shows an example of depth map in 2D gray format
impaired by compression distortion, and then be re-projected
in 3D point cloud format. In this example, structures in region
B is impaired, while discontinuity occurs in regions A and
C. The spread of FVV applications is limited if these kind of
distortions can not be healed properly.
In this literature, little work have been done on this special
distortion [3]–[5]. Some filters were proposed to address the
Fig. 1. The comparison of original and reconstructed depth map in 2D gray
format and 3D point cloud format
problem of depth map enhancement. For example, a joint
trilateral filter [6] and noise-aware depth filter [7] were respec-
tively proposed for in-loop filtering and suppress the texture-
copy defects. However, these methods are not specified for
depth maps with compression distortions [8]. Challenges arise
from two aspects. Actually, sharp edges and boundaries can
be hardly restored to their correction positions by conven-
tional non-joint filters, while compression distortion may be
propagated from color frames to depth maps if joint filters
are applied. In this paper, we propose a cluster-based cross-
view filter to address the problem of multi-view depth maps
impaired by compression distortions. We present our scheme
in details in Sec. II, and then the results and discussions in
Sec. III. Finally, we concluded our work in Sec. IV.
II. T P S
The proposed framework is illustrated in Fig. 2. In this
scheme, bilateral filter is modified and extended by cluster-
978-1-5090-5316-2/16/31.00
c
2016IEEE VCIP 2016, Nov. 27 – 30, 2016, Chengdu, China