The Marginal Importance of Distortions and Alignment in CASSI systems

TL;DR

This work tackles the gap between optical hardware and computational reconstruction in CASSI hyperspectral imaging by demonstrating that distortions and misalignments have only a marginal impact on reconstruction quality when a realistic propagation model is incorporated. It introduces a differentiable ray-tracing renderer within the dO framework to render realistic coded acquisitions for four CASSI configurations and five reconstruction algorithms, aided by a spatio-spectral mapping $f$ that initializes reconstructions. The study designs a fair comparison using a double-Amici prism and a single prism, validating the renderer against Zemax and confirming close PSF fidelity, and shows that reconstruction performance is dominated by the processing algorithm rather than the optical distortions. The contributions include end-to-end CASSI co-design capabilities, an accurate propagation-based initialization and mapping for reconstruction, and public release of code for DiffOptics and DiffCassiSim, enabling broader exploration of CASSI designs and information-theoretic assessments.

Abstract

This paper introduces a differentiable ray-tracing based model that incorporates aberrations and distortions to render realistic coded hyperspectral acquisitions using Coded-Aperture Spectral Snapshot Imagers (CASSI). CASSI systems can now be optimized in order to fulfill simultaneously several optical design constraints as well as processing constraints. Four comparable CASSI systems with varying degree of optical aberrations have been designed and modeled. The resulting rendered hyperspectral acquisitions from each of these systems are combined with five state-of-the-art hyperspectral cube reconstruction processes. These reconstruction processes encompass a mapping function created from each system's propagation model to account for distortions and aberrations during the reconstruction process. Our analyses show that if properly modeled, the effects of geometric distortions of the system and misalignments of the dispersive elements have a marginal impact on the overall quality of the reconstructed hyperspectral data cubes. Therefore, relaxing traditional constraints on measurement conformity and fidelity to the scene enables the development of novel imaging instruments, guided by performance metrics applied to the design or the processing of acquisitions. By providing a complete framework for design, simulation and evaluation, this work contributes to the optimization and exploration of new CASSI systems, and more generally to the computational imaging community.

Figures (7)

• Figure 1: Top-view of the systems layout for the (SP) configuration (a) and the (AP) configuration (b).
• Figure 2: Optical designs of dispersive elements in our imaging system. (a) A commercially available N-BK7 equilateral prism, aligned in for minimum deviation $D_0$. (b) A custom-designed double-Amici prism optimized to minimize distortions and achieve direct-view geometry. Both designs exhibit comparable angular spectral spreading $\Delta$ to ensure a fair comparison of system performance.
• Figure 3: Distortion maps of a regular grid of points traced through the four considered optical systems, for the extrema and center wavelengths (450 nm, 520 nm, and 650 nm). Note the different colorscales between the configurations.
• Figure 4: PSFs obtained with the misaligned single prism configuration (mSP) at four positions in the field of view (positions denoted by the red points on the bottom left of each figure). Top: PSFs obtained with dO, bottom: PSFs obtained with Zemax. The dotted black circle corresponds to the RMS radius centered on the centroid of the PSFs.
• Figure 5: Maps of the difference between distortions computed with dO and Zemax, for the extrema and central wavelengths (450 nm, 520 nm, and 650 nm). Note the same colorscales across the 4 configurations.