Adaptive compression for interferometric imaging spectroscopy data guided by fringe features

Abstract

To address the problems of high data redundancy and spectral information loss in spatiotemporally modulated interferometric imaging spectrometers under ultra low bit rate conditions, we present a proof-of-concept study of a fringe-aware adaptive sampling and reconstruction method based on physical feature perception of the interference field. Different from conventional universal compression algorithms, the proposed method fully exploits the nonuniform energy distribution characteristics of interferometric fringes and dynamically allocates sampling resources according to local fringe statistical features. Dense sampling is performed in high frequency fringe regions, especially near the zero optical path difference (OPD), to preserve critical spectral information. Meanwhile, a wavelet domain spectral compression framework combined with adaptive quantization is introduced to improve sparsity representation efficiency. During reconstruction, a KD-tree-based joint weighting model integrating spatial gradient consistency and spectral energy distribution constraints is employed to suppress oscillation artifacts and preserve spectral fidelity. Preliminary experimental results on the evaluated dataset demonstrate that the proposed method achieves promising rate–distortion performance under ultra low bit rate conditions. At a low compression rate of 0.10 bpp, the proposed method achieves a peak signal-to-noise ratio of 31.50 dB, a spectral angle mapper of 1.12 deg, and a Structural Similarity Index measure of 0.846, indicating its potential to preserve to preserve both spatial structures and spectral characteristics under highly compressed conditions. The proposed framework shows potential to provide a favorable rate-distortion trade off by exploiting fringe-aware adaptive sampling and spatial-spectral reconstruction. Preliminary ablation experiments further indicate that the fringe-aware adaptive sampling strategy substantially improves compression efficiency while maintaining reconstruction quality. These preliminary results indicate that the proposed method provides a promising proof-of-concept solution for potential real-time onboard compression of large-scale interferometric spectral data under resource-constrained conditions.