Paper Info
Title
Block-Level Entropy-Based Adaptive Sampling Framework for Depth Map
Abstract
Recently, the three dimensional (3-D) video technology has drawn significant attention among industry and academic researchers. As a special data format in 3-D video, the depth map consists of gray levels, which are nearly the same within an object but change abruptly across the boundaries. In view of the similarity of the gray levels in the most regions, down-sampling can be employed as the pre-processing and up-sampling as post-processing in most applications of the depth map, such as compression and transmission. Differently from the conventional uniform sampling, in this paper a framework of adaptive sampling is proposed based on block-level entropy according to the context of each block in the depth map. If more complicated context or more boundaries exist in the block, higher sampling rate should be set up while lower sampling rate should be used for smooth regions. Since the block-level entropy can represent the content complexity of each block in the depth map, it can be calculated to determine the adaptive sampling rate. In the experiments, different up-sampling methods are employed in our framework to test and verify the results. Experimental results show that compared with the uniform sampling, the proposed framework has higher objective and subjective quality at the same sampling rate both for the depth map and for the synthesized virtual viewpoint.
Year
DOI
Venue
2017
10.1109/BigMM.2017.18
2017 IEEE Third International Conference on Multimedia Big Data (BigMM)
Keywords
Field
DocType
Adaptive sampling,Block-level entropy,Depth map,3-D video
Iterative reconstruction,Computer vision,Data format,Adaptive sampling,Sampling (signal processing),Algorithm,Sampling (statistics),Artificial intelligence,Depth map,Distortion,Mathematics
Conference
ISBN
Citations 
PageRank 
978-1-5090-6550-9
0
0.34
References 
Authors
4
3
Name
Order
Citations
PageRank
Xiaoling Feng100.34
Bai Huihui224341.01
Yao Zhao31926219.11