Connecting Measured BRDFs to Analytic BRDFs by Data-Driven Diffuse-Specular Separation
Tiancheng Sun, Henrik Wann Jensen, Ravi Ramamoorthi
SIGGRAPH Asia 2018

We propose a novel framework for connecting measured and analytic BRDFs, by separating a measured BRDF into diffuse and specular components. This enables measured BRDF editing, a compact measured BRDF model, and insights in relating measured and analytic BRDFs. We also design a robust analytic fitting algorithm for two-lobe materials.


Three-dimensional Display via Multi-layer Translucencies
Tiancheng Sun, Huarong Gu

We built a display system using light field display and pyramid-like mirror. The displayed allows the observer to see the virtual 3D objects from different directions in the air.


Attribute‐preserving gamut mapping of measured BRDFs
Tiancheng Sun, Ana Serrano, Diego Gutierrez, Belen Masia
28th Eurographics Symposium on Rendering (EGSR 2017)

1st place at the SIGGRAPH 2017 ACM Student Research Competition (undergraduate category) SIGGRAPH annoucementSIGGRAPH blog postUCSD news
1st place at the 2018 ACM Student Research Competition (undergraduate category) ACM annoucementACM newsUCSD news

We proposed a new BRDF gamut mapping algorithm using two-step optimization. We optimize the luminance with the guide of perceptual attributes in the first step, and optimize the ink coefficients using image comparison in the second step.


Revisiting Cross-channel Information Transfer for Chromatic Aberration Correction
Tiancheng Sun, Yifan (Evan) Peng, Wolfgang Heidrich
2017 IEEE International Conference on Computer Vision (ICCV)

By modelling the similarity between different channels, we propose a new blind image deconvolution algorithm which transfers information from clear channel to blurry ones, and yield better results than state-of-the-art methods in both refractive and diffractive optical systems.


Convolution Neural Networks with Two Pathways for Image Style Recognition
Tiancheng Sun, Yulong Wang, Jian Yang, and Xiaolin Hu
IEEE Transactions on Image Processing (Volume: 26, Issue: 9, Sept. 2017)

We add the ideas used in image style transfer into traditional CNN network for image style recognition, and achieve state-of-the-art results on three benchmark datasets.