Haopeng Li

Haopeng Li received his Bachelor's degree from Southeast University in 2008. Then he received his Master's degree from KTH in 2010. He was a PhD student at Sound and Image Processing Lab at School of Electrical Engineering since 2010. He defended his Doctoral dissertation on Dec. 2015. His research interests include 3D rendering, video coding and mobile visual search.

Doctoral thesis:

Haopeng Li, “Feature-Based Image Processing for Rendering, Compression, and Visual Search”

Reseach interests

Mobile 3D Visual Search 

Mobile 3D visual search introduces the concept of 3D visual information into the search problem. It improves the search results by assessing the actual 3D geometry when compared to conventional appearance-based 2D image methods.

Database and Android app:

http://people.kth.se/~haopeng/M3DVS/index.html

Rate-constrained feature selection:

https://www.youtube.com/watch?v=FyXsGVpUYqQ

References:

• David Mars, Hanwei Wu, Haopeng Li, Markus Flierl, “Geometry-Based Ranking for Mobile 3D Visual Search using Hierarchically Structured Multi-View Features”, Proc. IEEE International Conference on Image Processing, Sept. 2015. [.pdf]

• David Mars, Hanwei Wu, Haopeng Li, Markus Flierl, “Joint Geometric Verification and Ranking using Multi-View Vocabulary Trees for Mobile 3D Visual Search”, Proc. Data Compression Conference, Apr. 2015. [.pdf]
• Xinrui Lyu, Haopeng Li, Markus Flierl, “Hierarchically Structured Multi-View Features for Mobile Visual Search”, Proc. Data Compression Conference, Mar. 2014. [.pdf]
• Haopeng Li, Markus Flierl, “Mobile 3D Visual Search using the Helmert Transformation of Stereo-Features”, Proc. IEEE International Conference on Image Processing, Sept. 2013. [.pdf]

Content-Adaptive Video Coding

The content-adaptive video coding divides the whole scene of the soccer video into dynamic and static content items. Each content item generates a subsequence from the input video. We exploit the different statistical properties of the static background and the dynamic content items to efficiently encode live soccer video.

References:

• Xiaohua Lu, Haopeng Li and Markus Flierl, “H.264-Compatible Coding of Background Soccer Video using Temporal Subbands”, Proc. IEEE Symposium on Multimedia, Dec. 2012. [.pdf]
• Haopeng Li and Markus Flierl, “Rate-Distortion-Optimized Content-Adaptive Coding for Immersive Networked Experience of Sports Events”, Proc. IEEE International Conference on Image Processing, Sept. 2011. [.pdf]

 

Image Feature-Based 3D Rendering, Modeling and Tracking

Reliable and sparse image features are used to extract the interview and inter-frame correlation among related views. Hence, accurate 3D information of image features can be efficiently utilized for real free-view rendering, object modeling and real time tracking.

Project webpage: http://www.projectfine.eu

References:

• Haopeng Li and Markus Flierl, “3D Model Hypotheses for Player Segmentation and Rendering in Free-Viewpoint Soccer Video”, Proc. IEEE Symposium on Multimedia, Dec. 2012. [.pdf]
• Haopeng Li and Markus Flierl, “SIFT-Based Modeling and Coding of Background Scenes for Multiview Soccer Video”, Proc. IEEE International Conference on Image Processing, Sept. 2012. [.pdf]
• Haopeng Li and Markus Flierl, “SIFT-Based Multi-View Cooperative Tracking for Soccer Video”, Proc. IEEE International Conference on Acoustics, Speech, and Signal Processing, Mar. 2012. [.pdf]

Multi-view cooperative tracking:

https://www.youtube.com/watch?v=vcOb6YUcYiA

Depth Image Enhancement

The depth image plays a crucial role for Depth Image Based Rendering (DIBR). However, most of the conventional depth image estimation approaches determine the depth
information from a limited set of nearby reference images. This leads to inconsistencies among multiple reference depth images, thus resulting in poor rendering quality.

References:

• Hannes Helgason, Haopeng Li and Markus Flierl, “Multiscale Framework for Adaptive and Robust Enhancement of Depth in Multi-View Imagery”, Proc. IEEE International Conference on Image Processing, Sept. 2012. [.pdf]
• Haopeng Li and Markus Flierl, “SIFT-Based Improvement of Depth Imagery”,
  Proc. IEEE International Conference on Multimedia and Expo, July 2011. [.pdf]