Design method of broadband filter array with high light efficiency

Abstract

Spectral filter array (SFA) imaging is a noval spectral image acquisition method that sets a filter array on one single sensor. It is widely used for environmental monitoring, agricultural assessment, geological exploration, and medical diagnosis. The spectral sensitivities on SFA highly influence the imaging and reconstruction quality. In low-light environments, the imaging quality of the narrowband filter array is significantly compromised, resulting in inadequate image brightness and loss of detail. To address the challenge that heuristic search methods often struggle to find the optimal solution, this paper presents a numerical design method for a global optimal broadband filter array with high light efficiency. The optimal broadband filter array is obtained by maximizing the minimal pairwise angles between filter vectors, and a light efficiency regularization term is used to ensure more light passes through. Further, a prototype spectral imaging system comprising a broadband filter array is built and demonstrated. The findings from the simulations conducted with publicly available multispectral datasets indicated that the broadband filter array demonstrated superior accuracy and enhanced noise resistance in the reconstruction of spectral information when compared to the narrowband filter array. Simultaneously, the simulation results demonstrated that excessive light efficiency will reduce both the anti-noise ability and the accuracy of the reconstruction. Thus, it is essential to choose a broadband filter array that demonstrates satisfactory light efficiency. This study will be beneficial to optimizing the filter array of broadband spectral imaging sensors.