computational imaging

High bit-depth projection
A new approach for achieving high bit-depth projection by modulating the intensity of the projector's light source

We demonstrate a novel structured light system that uses only a line-sensor to obtain high-resolution 3D scans of a scene.

White balancing
We demonstrate a per-pixel white balancing algorithm for Lambertian objects using flash photography. The technique works with arbitrarily complex lighting.

Lensfree cameras
We demonstrate a thin form-factor camera design that consists of a bare-sensor with a mask places a few 100 micrometers in front

A novel architecture for video CS that optically resolves scenes at high-spatial resolution using a line-detector with just 1000px

Novel architecture for compressive imaging in short-wave infrared

Compressive epsilon photography
Complete freedom in changing the aperture and focus settings from a stack of photographs

High-speed imaging
Converting a low-fps device into a high-speed camera using a global shutter

Removing blur due to hand-shake from multiple images obtained in burst

compressive sensing

Compressive random features
We tighten the performance gaurantees for random features when applied on sparse data. We also show that random features can be combined with compressive measurements.

Cross-scale predictive dictionaries
We enable multi-scale sparse representations that are endowed with cross-scale predictive structures, similar to wavelet tree models.

Learning deterministic, near-isometric embeddings using semi-definite programming

Video compressive sensing with the single-pixel camera using motion compensation and dual-scale sensing matrices

Union-of-subspace learning
Performance gaurantees for union-of-subspace recovery using OMP and endogenous sparse coding

Direct estimation of the parameters of a linear dynamical system from compressive measurements

A greedy approach to estimating a low-rank and sparse components of a matrix from compressive measurements

Compressive background subtraction
Direct recovery of foreground objects from compressive measurements

shape and reflectance

Shape and reflectance from two-bounce light
We show that shape and reflectance of non-convex objects can be reliably estimated purely from two-bounce light transients.

Small baseline photometric stereo
Analysis of photometric stereo for SWaP-constrained acquisition

Transparent shape recovery
Quantifying the information encoded in a single light-ray to perform single view-based shape estimation of transparent objects

Spatially-varying reflectance
Exemplar-based photometric stereo for per-pixel shape and reflectance estimation

Image invariants for mirrors
We derive a photometric invariant for smooth reflective mirrors and show that image gradients at parabolic curvature points are constrained

Shape recovery of mirrors
Recovering shape of mirrors under a locally quadric shape model and sparse image correspondences