I. Introduction
Depth and spectrum are two different dimensions that are generally missing in ordinary digital imaging, yet both provide indispensable information for interpreting the scene. Originally, either depth or spectral imaging requires a scanning strategy for accurate measurement, which is unfavorable in time-critical applications. Recently, owing to the rapid advancement of image sensors and computing power, snapshot depth and spectral cameras have been developed separately with improved resolution and accuracy. Specifically, for depth imaging, techniques such as stereo matching [1], [2], structured light [3], [4], and time-of-flight [5], [6] can all generate decent video-rate depth maps; for spectral imaging, snapshot systems such as compute tomographic imaging spectrometry [7], [8], prism-mask multispectral imaging system [9], [10], and coded aperture snapshot spectral imager (CASSI) [11], [12] all support reliable capturing of dynamic scenes through computational reconstruction. It can be predicted that, just as Microsoft Kinect has made depth cameras an emerging new configuration for digital imaging, ubiquitous spectral cameras will also be seen in the near future.