Data compression for snapshot mosaic hyperspectral image sensors

Recent achievements in hyperspectral imaging (HSI) demonstrated successfully a novel snapshot mosaic sensor architecture, enabling spectral imaging in a truly compact way. Integration of this new technology in handheld devices necessitates efficient compression of HSI data. However, due to the specific mosaic structure of the acquired images, traditional compression methods tailored to full-resolution HSI data cubes fail to exploit the special spatio-spectral interrelations among the pixels. This paper introduces an efficient and computationally tractable compression technique for mosaic HSI images. Specifically, an appropriate decorrelator is constructed for exploiting the spatio-spectral redundancies among the pixels, by modeling the filters arrangement on the mosaic HSI sensor as a multiple-input multiple-output antenna array. Doing so, the decorrelator depends only on the sensor and not on the data to be compressed. Comparison with state-of-the-art compression methods designed for HSI data cubes reveals that our approach achieves better reconstruction quality at lower bits-per-pixel rates.