Adaptive Parameters and Edge Point Guided Depth Image Super-resolution
[1]Wan Wenbo,Zu Lanjing,Xue Zeying,et al.Adaptive Parameters and Edge Point Guided Depth Image Super-resolution[J].Journal of Zhengzhou University (Engineering Science),2021,42(03):33-.[doi:10.13705/j.issn.1671-6833.2021.03.006]
Copy
Journal of Zhengzhou University (Engineering Science)[ISSN
1671-6833/CN
41-1339/T] Volume:
42卷
Number of periods:
2021 03
Page number:
33-
Column:
Public date:
2021-05-10
- Title:
- Adaptive Parameters and Edge Point Guided Depth Image Super-resolution
- Keywords:
- depth image; super-resolution reconstruction; adaptive parameters; edge point detection; depth correction
- CLC:
- -
- DOI:
- 10.13705/j.issn.1671-6833.2021.03.006
- Abstract:
- Depth cameras such as TOF can obtain low-resolution depth images only, and cannot meet the needs of 3D vision systems. High-resolution depth images can be obtained by the super-resolution algorithm of depth images, but with problems such as artifacts in local areas of the image and unclear edge structures due to texture replication. Based on the HR depth pseudo-matrix, an algorithm of depth value reconstruction guided by adaptive parameters and edge points is proposed. In this paper, the pixels in the edge area of the low-resolution depth image are used to find the pixels with the wrong depth value in the correction image: and clustering is performed to obtain the reconstructed depth image. By introducing an autoregressive model guided by adaptive parameters, the depth value of pixels in the edge area is predicted. The experimental results prove that the algorithm can effectively reduce the ambiguity of the depth discontinuous area, and obtain higher quality high-resolution depth images. Compared with the existing algorithms, when the upsampling factors are 2, 4 and 8, the average dead pixel rate of the output results of this algorithm is lower than 0.1, which can effectively verify the advantages of the algorithm in this paper.
- References:
-
[1] REN Y, LIU J, YUAN H, et al. Depth up-sampling via pixel-classifying and joint bilateral filtering[J]. KSII transactions on internet & information systems, 2018, 12(7):3217-3238.[2] GU S, ZUO W, GUO S, et al. Learning dynamic guidance for depth image enhancement[C]// 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). Piscataway:IEEE, 2017:712-721.[3] YANG J Y, YE X C, LI K, et al. Color-guided depth recovery from RGB-D data using an adaptive autogressive model [J]. IEEE transactions on image processing, 2014, 23(8): 3443-3458.[4] BAE S, PARIS S, DURAND F. Two-scale tone management for photographic look [J]. ACM transactions on graphics, 2006, 25(3): 637-645.[5] QIAO Y G, JIAO L C, YANG S Y, et al. A novel segmentation based depth map up-sampling[J]. IEEE transactions on multimedia, 2019, 21(1): 1-14.[6] FERSTL D, REINBACHER C, RANFTL R, et al. Image guided depth upsampling using anisotropic total generalized variation[C]//Proceedings of the 2013 IEEE International Conference on Computer Vision. New York:IEEE, 2013:993-1000.[7] WEN Y, SHENG B, LI P, et al. Deep color guided coarse-to-fine convolutional network cascade for depth image super-resolution[J]. IEEE Transactions on Image Processing, 2019, 28(2): 994-1006.[8] Middlebury datasets [EB/OL]. (2001-02-11)[2020-09-15].http://vision.middlebury.edu/stereo/data/.[9] 胡春鹤,王依帆,朱书豪,等.基于鸽群优化算法的图像分割方法研究[J].郑州大学学报(工学版),2019,40(4):42-47.
- Similar References:
Memo
-
Last Update:
2021-06-24