Research Projects

Generative Image Composite

Computational Photography, Machine Learning

Oct, 2016 — Present, 2016


Photometric Consistency for Fast-forward Videos

Computational Photography, Image Processing

May, 2016 — Nov, 2016

My summer intern project at Adobe Imagination Lab

Vision Correcting Light Field Display

Computational Display, Computer Graphics

Sep, 2015 — Present

rose_new50mm_5mmPinHole_128x128_5x5This is a joint project between me and Jinkyu Kim.

Wearable optical corrections (glasses and contact lenses) have been the primary option but are prone to inconvenience. Imagine a person with farsightedness needs to put on glasses whenever he/she wants to check the time on a phone. The goal of this project is to create a vision correcting display that compensates for eye aberration and allows for continuous eye movement within a viewing zone. (the image on the left is actually a light field image :-))

We are working on a paper submission and more info would be updated soon.

Virtual Fitting Room

Computer Graphics, Machine Learning

May, 2013 — Dec, 2014


This is a project I proposed with my partner Lam Yuk. We spent a whole year on prototyping our virtual fitting room, a system that helps online clothes shoppers to virtually try on clothes they want to buy. This project appears to be one of the most successful and my favorite projects so far. Our short paper is accepted by CASA Conference 2014 and the poster is selected to enter ACM SIGGRAPH 2014 Student Research Competition. We also won the grand prize at Rice’s design showcase for innovative design.

The short paper can be viewed here

The poster can be viewed here

Fourier Ptychography

Fourier Optics, Computational Photography, Computer Vision

Mar, 2014 — May, 2015

In microscopic applications, wide-field and high resolution images are desired. However, there is always a trade off between magnification factor and the field of view. In this project, a technique called Fourier Ptychographic Microscopy is used to overcome the problem. Briefly speaking, the key is to stitch a number low resolution images in Fourier space to recover an accurate high-resolution image. A main purpose of the project is to build a smartphone based portable microscopic device that utilizes ptychography to generate wide-field and high resolution images.

Current Status:

  • Implemented ptychography simulation that can retrieve phase and amplitude information of the target high-resolution image
  • Designed a multiplexing strategy to reduce image-acquisition time

Pre-rotation effect on Image Coding with Contourlet Transform

Image and Signal Processing

May, 2013 — Aug, 2013

One disadvantage of Image Compression with Wavelet is its limited directional sensitivity, as for most time edges are not aligned horizontally or vertically. Contourlet Transform overcomes the problem by offering a higher directional sensitivity. However, energy of an image can still hardly concentrate if edges fall between two filtering directions. My research is to search for an optimal direction an image can be pre-roated to so that edges, or most its energy can be encoded during compression. In this way, higher compression ratio together with higher SNR can be achieved.

Results and Conclusions:

  • three-pass-shearing is the best rotation method in this application
  • compression performances depend on the rotation angle of the image