menu

Learning to Synthesize Motion Blur Fall 2018

CVPR, 2019 (Oral Presentation)

It is difficult to portray a sense of movement in a single image. We present a novel technique to synthesize motion blur, which creates a visual effect to summarize movement, such as portraying a car racing by or the commotion of a busy city intersection. It is also useful to temporally smooth timelapse videos and rendered animations.

Unprocessing Images for Learned Raw Denoising Fall 2018

CVPR, 2019 (Oral Presentation)

Photographs often exhibit noise, especially in low light. While denoise neural networks work well on synthetic inputs, they often fail on real noisy images. We present the new approach to "unprocess" images, which creates more realistic training data and produces state-of-the-art results on real photos.

Dancing Music Box Spring 2016

Finally! The product we've all been waiting for: a magic box that makes objects dance to music! Funded by Carnegie Mellon's Build18 festival, the project is motivated by the KickStarter campaign Slow Dance – A Frame that Slows Down Time, with the added goal to time the resulting movement to the rhythm of any song.

HDR+ Pipeline Implementation Fall 2016

Inspired by Google's recent Burst Photography publication, this technique combines multiple, underexposed raw frames as a means of noise removal, and later applies tone mapping to maintain local contrast while brightening shadows and darkening highlights.

Scotty3D Stereoscopic Renderer Fall 2016

Expanding on the Scotty3D renderer, we supplement the ray tracing segment of this application with stereo rendering, targeted for headsets such as the Oculus Rift and Google Cardboard. The project effectively adds dual camera support, and inverts the pincushion distortion caused by the headset lenses.

Eye Tracking Photo Refocusing Fall 2016

This research project, funded by Carnegie Mellon University's Crosswalk grant, aims to allow photos to dynamically change focus based on where in the image a viewer is looking. This new display will provide photographers with another compositional dimension to consider in their work, and may enable photos to tell a more dynamic or immersive story.

Parallel Seam Carving Approximation Spring 2016

To significantly improve the runtime of the traditional dynamic programming seam carving algorithm, we devised an approximation that provides similar output images while relieving data dependencies, and implemented a parallelized version of the approximation algorithm, which demonstrates superior speed.