Ultrafast sources and detectors have been used to record the time-resolved scattering of light propagating through macroscopic scenes. In the context of computational imaging, decomposition of this transient light transport (TLT) is useful for applications, such as characterizing materials, imaging through diffuser layers, and relighting scenes dynamically. Here, we demonstrate a method of convolutional sparse coding to decompose TLT into direct reflections, inter-reflections, and subsurface scattering. The method relies on the sparsity composition of the time-resolved kernel. We show that it is robust and accurate to noise during the acquisition process.
© 2014 Optical Society of America
Fourier Optics and Signal Processing
Original Manuscript: January 16, 2014
Revised Manuscript: April 14, 2014
Manuscript Accepted: April 16, 2014
Published: May 23, 2014
Xuemei Hu, Yue Deng, Xing Lin, Jinli Suo, Qionghai Dai, Christopher Barsi, and Ramesh Raskar, "Robust and accurate transient light transport decomposition via convolutional sparse coding," Opt. Lett. 39, 3177-3180 (2014)