Kaihuai Qin
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Kaihuai Qin has a total of 29 co-authors. Here are the top co-authors by number of shared publications:
- Chong Zhao
Department of Computer Science and Engineering, The Chinese University of Hong Kong, Hong Kong
1 shared publication Denggao Li
Tsinghua University
1 shared publication- Hanqiu Sun
Department of Computer Science and Engineering, The Chinese University of Hong Kong, Hong Kong
1 shared publication - Hua-Wei Wang
Department of Computer Science and Technology, Tsinghua University, Beijing, PR China
1 shared publication - Hua-Wei Wang
Tsinghua University
1 shared publication Huawei Wang
Tsinghua University
1 shared publication
Latest Publications
- Open Access
Unsupervised High-Resolution Depth Learning From Videos With Dual Networks
Unsupervised depth learning takes the appearance difference between a target view and a view synthesized from its adjacent frame as supervisory signal. Since the supervisory signal only comes from images themselves, the resolution of training data significantly impacts the performance. High-resolution images contain more fine-grained details and provide more accurate supervisory signal. However, due to the limitation of memory and computation power, the original images are typically down-sampled during training, which suffers heavy loss of details and disparity accuracy. In order to fully explore the information contained in high-resolution data, we propose a simple yet effective dual networks architecture, which can directly take high-resolution images as input and generate high-resolution and high-accuracy depth map efficiently. We also propose a Self-assembled Attention (SA-Attention) module to handle low-texture region. The evaluation on the benchmark KITTI and Make3D datasets demonstrates that our method achieves state-of-the-art results in the monocular depth estimation task.
Open AccessUnsupervised High-Resolution Depth Learning From Videos With Dual Networks
Unsupervised depth learning takes the appearance difference between a target view and a view synthesized from its adjacent frame as supervisory signal. Since the supervisory signal only comes from images themselves, the resolution of training data significantly impacts the performance. High-resolution images contain more fine-grained details and provide more accurate supervisory signal. However, due to the limitation of memory and computation power, the original images are typically down-sampled during training, which suffers heavy loss of details and disparity accuracy. In order to fully explore the information contained in high-resolution data, we propose a simple yet effective dual networks architecture, which can directly take high-resolution images as input and generate high-resolution and high-accuracy depth map efficiently. We also propose a Self-assembled Attention (SA-Attention) module to handle low-texture region. The evaluation on the benchmark KITTI and Make3D datasets demonstrates that our method achieves state-of-the-art results in the monocular depth estimation task.
- 14 September 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computational Imaging
- Vol. 4 (4) , 632-639
Fast 3-D Ultrasonic Imaging Using Time-Domain Synthetic Aperture Focusing Techniques Based on Circular Scan Conversions
DAB-SAFT (Drawing-Arc-Based Synthetic Aperture Focusing Technique) is capable of greatly improving the imaging speed by reinterpreting the focusing process in the forward direction, but it is only suitable for 2D ultrasonic testing. By extending DAB-SAFT to fast 3D ultrasonic imaging, this paper proposes both spherical scan-conversion-based SAFT and a Spherical-Surface-Conversion-Based SAFT (SSCB-SAFT for short) for 3D non-destructive testing (NDT). In SSCB-SAFT, the entire imaging procedure is converted to multiple scan-conversion operations of spherical surfaces. To obtain the coordinates of pixels on a given spherical surface, the spherical scan-conversion algorithm is presented based on circular scan conversions to speed up the imaging procedure by avoiding the time-consuming computations of root-mean-square distances. Besides, the performance of DAB-SAFT is further improved based on the relative positions between different scanning positions, which largely reduces the number of circular arc scan-conversion operations. This acceleration strategy is also applicable to SSCB-SAFT. The simulation experiments show that the improved SSCB-SAFT speeds up the imaging procedure 4 times while maintaining the same results.
- 14 September 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Computational Imaging
- Vol. 4 (4) , 632-639
Fast 3-D Ultrasonic Imaging Using Time-Domain Synthetic Aperture Focusing Techniques Based on Circular Scan Conversions
DAB-SAFT (Drawing-Arc-Based Synthetic Aperture Focusing Technique) is capable of greatly improving the imaging speed by reinterpreting the focusing process in the forward direction, but it is only suitable for 2D ultrasonic testing. By extending DAB-SAFT to fast 3D ultrasonic imaging, this paper proposes both spherical scan-conversion-based SAFT and a Spherical-Surface-Conversion-Based SAFT (SSCB-SAFT for short) for 3D non-destructive testing (NDT). In SSCB-SAFT, the entire imaging procedure is converted to multiple scan-conversion operations of spherical surfaces. To obtain the coordinates of pixels on a given spherical surface, the spherical scan-conversion algorithm is presented based on circular scan conversions to speed up the imaging procedure by avoiding the time-consuming computations of root-mean-square distances. Besides, the performance of DAB-SAFT is further improved based on the relative positions between different scanning positions, which largely reduces the number of circular arc scan-conversion operations. This acceleration strategy is also applicable to SSCB-SAFT. The simulation experiments show that the improved SSCB-SAFT speeds up the imaging procedure 4 times while maintaining the same results.
- 11 June 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Visualization and Computer Graphics
- Vol. 25 (7) , 2392-2403
Biorthogonal Wavelet Transforms and Applications Based on Generalized Progressive Catmull-Clark Subdivision with Shape Control
In recent years, some biorthogonal Catmull-Clark subdivision wavelets constructed via lifting scheme have been proposed to speed up processing of geometric models. Thanks to the idea of progressive interpolation, the compression qualities and noise-filtering effects have been improved significantly. However, the reconstruction precision fails to be improved further because many model details are removed and the noise-filtering performance decreases greatly while the noise intensity increases gradually. To deal with this dilemma, a unified Catmull-Clark subdivision based biorthogonal wavelet construction with shape control parameters is presented to process 3D models with sharp-feature constraints. By customizing its local orthogonalizing coefficients for different vertex valences of quadrilateral patches, the novel scheme can greatly strengthen the capability of the model's shape control that is vital for data compression, noise-filtering, etc. Combined with the local and in-place lifting operations, the proposed wavelet transform can dramatically decrease the memory consumption and computation complexity. Both theoretical analysis and experiments show that, compared with the state-of-the-art lifting-based solutions, the proposed wavelet transform achieves higher compression ratio, more stable noise-filtering effects and better progressive transmission quality, not only decreasing Bits/vertex of 3D meshes and improving PSNR of reconstruction models, but also reducing the time costs of coding and decoding.
- 11 June 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Visualization and Computer Graphics
- Vol. 25 (7) , 2392-2403
Biorthogonal Wavelet Transforms and Applications Based on Generalized Progressive Catmull-Clark Subdivision with Shape Control
In recent years, some biorthogonal Catmull-Clark subdivision wavelets constructed via lifting scheme have been proposed to speed up processing of geometric models. Thanks to the idea of progressive interpolation, the compression qualities and noise-filtering effects have been improved significantly. However, the reconstruction precision fails to be improved further because many model details are removed and the noise-filtering performance decreases greatly while the noise intensity increases gradually. To deal with this dilemma, a unified Catmull-Clark subdivision based biorthogonal wavelet construction with shape control parameters is presented to process 3D models with sharp-feature constraints. By customizing its local orthogonalizing coefficients for different vertex valences of quadrilateral patches, the novel scheme can greatly strengthen the capability of the model's shape control that is vital for data compression, noise-filtering, etc. Combined with the local and in-place lifting operations, the proposed wavelet transform can dramatically decrease the memory consumption and computation complexity. Both theoretical analysis and experiments show that, compared with the state-of-the-art lifting-based solutions, the proposed wavelet transform achieves higher compression ratio, more stable noise-filtering effects and better progressive transmission quality, not only decreasing Bits/vertex of 3D meshes and improving PSNR of reconstruction models, but also reducing the time costs of coding and decoding.
- 20 January 2011
- Published by China Science Publishing & Media Ltd. in Journal of Computer Applications
- Vol. 31 (1) , 65-69
Self-feedback photometric correction approach for multi-projector tiled display
- 20 January 2011
- Published by China Science Publishing & Media Ltd. in Journal of Computer Applications
- Vol. 31 (1) , 65-69
Self-feedback photometric correction approach for multi-projector tiled display
- 17 April 2010
- Published by Springer Nature in The Visual Computer
- Vol. 26 (6-8) , 457-465
GPU rendering for tiled multi-projector autostereoscopic display based on chromium
- 17 April 2010
- Published by Springer Nature in The Visual Computer
- Vol. 26 (6-8) , 457-465
GPU rendering for tiled multi-projector autostereoscopic display based on chromium
- Open Access
- 28 September 2009
- Published by MDPI AG in Algorithms
- Vol. 2 (4) , 1263-1280
Compound Biorthogonal Wavelets on Quadrilaterals and Polar Structures
In geometric models with high-valence vertices, current subdivision wavelets may not deal with the special cases well for good visual effect of multiresolution surfaces. In this paper, we present the novel biorthogonal polar subdivision wavelets, which can efficiently perform wavelet analysis to the control nets with polar structures. The polar subdivision can generate more natural subdivision surfaces around the high-valence vertices and avoid the ripples and saddle points where Catmull-Clark subdivision may produce. Based on polar subdivision, our wavelet scheme supports special operations on the polar structures, especially suitable to models with many facets joining. For seamless fusing with Catmull-Clark subdivision wavelet, we construct the wavelets in circular and radial layers of polar structures, so can combine the subdivision wavelets smoothly for composite models formed by quadrilaterals and polar structures. The computations of wavelet analysis and synthesis are highly efficient and fully in-place. The experimental results have confirmed the stability of our proposed approach.
- 28 September 2009
- Published by MDPI AG in Algorithms
- Vol. 2 (4) , 1263-1280
Open AccessCompound Biorthogonal Wavelets on Quadrilaterals and Polar Structures
In geometric models with high-valence vertices, current subdivision wavelets may not deal with the special cases well for good visual effect of multiresolution surfaces. In this paper, we present the novel biorthogonal polar subdivision wavelets, which can efficiently perform wavelet analysis to the control nets with polar structures. The polar subdivision can generate more natural subdivision surfaces around the high-valence vertices and avoid the ripples and saddle points where Catmull-Clark subdivision may produce. Based on polar subdivision, our wavelet scheme supports special operations on the polar structures, especially suitable to models with many facets joining. For seamless fusing with Catmull-Clark subdivision wavelet, we construct the wavelets in circular and radial layers of polar structures, so can combine the subdivision wavelets smoothly for composite models formed by quadrilaterals and polar structures. The computations of wavelet analysis and synthesis are highly efficient and fully in-place. The experimental results have confirmed the stability of our proposed approach.
- 1 December 2008
- Published by Elsevier in Computer Aided Geometric Design
- Vol. 25 (9) , 816-836
Biorthogonal wavelets based on gradual subdivision of quadrilateral meshes
- 1 December 2008
- Published by Elsevier in Computer Aided Geometric Design
- Vol. 25 (9) , 816-836
Biorthogonal wavelets based on gradual subdivision of quadrilateral meshes
- 1 March 2006
- Published by Springer Nature in Journal of Computer Science and Technology
- Vol. 21 (2) , 166-172
Reflection and Refraction on Implicit Surfaces
- 1 March 2006
- Published by Springer Nature in Journal of Computer Science and Technology
- Vol. 21 (2) , 166-172
Reflection and Refraction on Implicit Surfaces
- 1 February 2005
- Published by Springer Nature in The Visual Computer
- Vol. 21 (1-2) , 59-70
Surface modeling with ternary interpolating subdivision
- 1 February 2005
- Published by Springer Nature in The Visual Computer
- Vol. 21 (1-2) , 59-70
Surface modeling with ternary interpolating subdivision
- 1 September 2004
- Published by Springer Nature in Journal of Computer Science and Technology
- Vol. 19 (5) , 657-664
Estimating subdivision depth of Catmull-Clark surfaces
- 1 September 2004
- Published by Springer Nature in Journal of Computer Science and Technology
- Vol. 19 (5) , 657-664
Estimating subdivision depth of Catmull-Clark surfaces