Summary: we present a system named AutoLF for analyzing lymph functions using NIR fluorescence images. In order to reduce the manual labor and improve the reliability of the measurement, we develop a number of image processing algorithms, including an object tracking algorithm to stabilize the subject, an image representation named flow map to characterize lymph flow, and a fully automatic algorithm to compute lymph velocity and frequency of propulsion.  

Jingdan Zhang, S. Kevin Zhou, Xiaoyan Xiang, John C. Rasmussen, and Eva M. Sevick-Muraca. An Image Analysis System for Near-infrared (NIR) Fluorescence Lymph Imaging. SPIE Medical Imaging, 2011.(pdf)

John C. Rasmussen, Merrick Bautista, I-Chih Tan, Kristen E. Adams, Melissa Aldrich, Milton V. Marshall, Caroline E. Fife, Eric A. Maus, Latisha A. Smith, Jingdan Zhang, Xiaoyan Xiang, S. Kevin Zhou, and Eva M. Sevick-Muraca, Validation of ALFIA: a platform for quantifying near-infrared fluorescent images of lymphatic propulsion in humans, SPIE Biomedical Optics, 2011.(pdf)

Summary: We propose a fully automatic system to detect cysts jointly in both B-mode and elasticity images. It is based on database-guided techniques that learn the knowledge of cyst appearance automatically from B-mode and elasticity images in a database. Further, for a detected cyst in a query image, the cysts with similar image appearance in the database are retrieved to improve diagnostic accuracy and confidence. 

Jingdan Zhang, S. Kevin Zhou, Shelby Brunke, Carol Lowery, and Dorin Comaniciu. Detection and Retrieval of Cysts in Joint Ultrasound B-Mode and Elasticity Breast Images. IEEE International Symposium on Biomedical Imaging (ISBI), 2010. (pdf)

Summary: We propose a fully automatic system to detect and segment breast tumors in 2D ultrasonography. For tumor detection, we apply a classification approach to discriminate between tumors and their background. For tumor segmentation, we propose a discriminative graph cut algorithm, where the data fidelity function is online learned and the data compatibility function is offline learned, both discriminatively. We demonstrate the performance of the proposed algorithms on a large image database of breast tumors. 

Jingdan Zhang, S. Kevin Zhou, Shelby Brunke, Carol Lowery, and Dorin Comaniciu. Database-Guided Breast Tumor Detection and Segmentation in 2D Ultrasound Images. SPIE Medical Imaging, 2010. (pdf)

Summary: We presented Detection in a Cloud (DiC) system for anatomical landmark detection in the cloud computing environment. At the core of the system is a hierarchical learning algorithm that propagates position candidate hypotheses across a hierarchy of classifiers during training and detection. The total bandwidth savings for retrieving remotely stored data amount to 50 times (CT data) and 300 times (MRI data) reduction when compared to the original data size and 4.5 times (CT) and 11.5 (MRI) when compared to data size after lossless compression. 

Michal Sofka, Kristof Ralovich, Jingdan Zhang, and S. Kevin Zhou, and Dorin Comaniciu. Progressive Data Transmission for Hierarchical Detection in a Cloud. The 2nd International Workshop on Medical Image Computing for Image-Assisted Clinical Intervention and Decision-Making (HP-MICCAI), 2010, Best Paper Award.(pdf)  

Summary: We present a Sequential Monte Carlo based Hierarchical Detection Network (HDN) for detecting multiple objects. The order of detection is automatically determined by a greedy algorithm that puts the most reliable detections earlier in the detection sequence. The detectors are organized in a multi-scale hierarchy with the scale parameter included in the order selection process.  

Michal Sofka, Jingdan Zhang, S. Kevin Zhou, and Dorin Comaniciu. Multiple Object Detection by Sequential Monte Carlo and Hierarchical Detection Network. IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2010.(pdf)  

Michal Sofka, Kristof Ralovich, Neil Birkbeck, Jigndan Zhang, S.Kevin Zhou. Integrated Detection Network (IDN) for Pose and Boundary Estimation in Medical Images. IEEE International Symposium on Biomedical Imaging (ISBI), 2011.(pdf)

Summary: We present a comparative study on how to apply three discriminative learning approaches - classification, regression, and ranking - to deformable shape segmentation. By casting the segmentation into a discriminative learning framework, a target fitting function can be steered to possess a desired shape for ease of optimization yet better characterize the relationship between shape and appearance. 

Jingdan Zhang, S. Kevin Zhou, Dorin Comaniciu and Leonard McMillan. Discriminative Learning for Deformable Shape Segmentation: A Comparative Study. ECCV 2008.(pdf)

Jingdan Zhang, S. Kevin Zhou, Leonard McMillan and Dorin Comaniciu. Joint Real-time Object Detection and Pose Estimation Using Probabilistic Boosting Network. CVPR 2007.(pdf)

Summary: Illumination inconsistencies cause serious problems for classical computer vision applications. We present a new approach to model illumination variations using an Illumination Ratio Map (IRM). The IRM is modeled as Markov network and be easily incorporated into low-level vision problems, such as tracking and stereo matching.

Jingdan Zhang, Jingyi Yu and Leonard McMillan. Robust Tracking and Stereo Matching under Variable Illumination.  CVPR, 2006. (pdf)

Summary: Motion capture data from human subjects exhibits considerable redundancy. We exploit this  redundancy by representing Mocap data with piecewise local linear components, which are determined via a divisive clustering method. This technique can be used to predict the complete configuration of a human model based on a subset of Mocap information as well as compressing and indexing motion databases.

Guodong Liu, Jingdan Zhang, Wei Wang and Leonard McMillan. A system for analyzing and indexing human motion databases (demo). Proc. ACM SIGMOD International Conference on Management of Data (SIGMOD), 924-926, 2005. (pdf) 

Guodong Liu, Jingdan Zhang, Wei Wang and Leonard McMillan. Human Motion Estimation from a Reduced Marker Set. To appear ACM SIGGRAPH Symposium on Interactive 3D Graphics (I3D), 2006.(pdf)

Progressively-Variant Texture Synthesis

Summary: We present an approach for decorating surfaces with progressively-variant textures. A progressively-variant texture can model local texture variations, including the scale, orientation, color, and shape variations of texture elements. We developed techniques for modeling progressively-variant textures in 2D as well as for synthesizing them over surfaces.

Jingdan Zhang, Kun Zhou, Luiz Velho, Baining Guo and Heung-Yeung Shum. Synthesis of Progressively-Variant Textures on Arbitrary Surfaces. ACM Transactions on Graphics(Proc. ACM SIGGRAPH), 295-302, 2003. (pdf)

Summary: The bidirectional texture function (BTF) is a 6D function that can describe textures arising from both spatially-variant surface re-flectance and surface mesostructures. For both BTF synthesis and hardware-accelerated rendering, a main challenge is handling the large amount of data in a BTF sample. In this project, we developed algorithms to synthesize BTF to arbitrary surfaces and render the synthesized BTF with GPU acceleration.

Xin Tong, Jingdan Zhang, Ligang Liu, Xi Wang, Baining Guo and Heung-Yeung Shum. Synthesis of Bidirectional Texture Functions on Arbitrary Surfaces. ACM Transactions on Graphics(Proc. ACM SIGGRAPH), 665-672, 2002. (pdf)

Xinguo Liu, Yaohua Hu, Jingdan Zhang, Xin Tong, Baining Guo and Heung-Yeung Shum. Synthesis and Rendering of Bidirectional Texture Functions on Arbitrary Surfaces. IEEE Transactions on Visualization and Computer Graphics, 10(3): 278-289, 2004.(pdf)

Summary: This project is related to my master thesis. I have built a prototype system that can automatically reconstruct the geometric as well as texture information from real objects. In this system, a four-freedom robot controls the position and orientation of an object and I use stereo vision and structured light to recovery geometric information of  the object.

Jingdan Zhang, Realistic Modeling Techniques Based On Real-World Sampling Dataset, Master's degree thesis, June. 2003. (Abstract) (pdf, Chinese) 

Three-Dimensional Biomedical Image Interpolation

Summary: We present a novel three-dimensional gray-level interpolation method called Directional Coherence Interpolation (DCI). DCI interpolates the missing image data along the maximum coherence directions (MCD), which are estimated from the local image intensity yet constrained by a generic smoothness term. The principal advantage of the proposed approach is that it leads to significantly higher visual quality in 3D rendering when compared with traditional biomedical image interpolation methods.

Jingdan Zhang, Yongmei Wang and Baining Guo. Pyramidal Search of Maximum Coherence Direction for Biomedical Image Interpolation. IEEE International Symposium on Biomedical Imaging, 887-890, 2002.(pdf)

Yongmei Michelle Wang, Jingdan Zhang, Zhunping Zhang, Baining Guo. Directional Coherence Interpolation for Three-Dimensional Gray-Level Images. International Journal of Image and Graphics, 4(4), 535-561, 2004.(pdf)

Image Segmentation for image retrieval system

Summary: We propose a algorithm efficiently combining the local and global information to achieve unsupervised segmentation of color images. 

Jingdan Zhang, Zhidong Deng, Baining Guo. Two Stage Unsupervised Segmentation of Color Images. Proc. Chinagraph, 144-148, Beijing, Sept 2002. (Abstract) (pdf, Chinese)

High quality texture mapping

Summary: We describe a new method to map a texture on a surface with a spatially-variant filter. Our filter takes into consideration the effects of anisotropy using a Jacobian approximation while computing the sampling rate, and the interpolation weights are computed with a sinc function.

Ke Deng, Jingdan Zhang, Lifeng Wang and Baining Guo. Texture Mapping with a Jacobian-Based Spatially-Variant Filter. Proc. IEEE Pacific Graphics, 2002.(pdf)