A Novel Framework for Mixed Noise Removal From Greenhouse Gases Monitoring Instrument (GMI) Interferogram Images on GF5-02 Satellite

To cope with global climate change and monitor the concentration distribution of CO2 and CH4 in the troposphere, the greenhouse gases monitoring instrument (GMI) carried by the GF5-02 satellite was successfully launched on September 7, 2021. GMI utilizes spatial heterodyne spectroscopy techniques to acquire interferogram images. During the acquisition process, interferogram images are inevitably corrupted by various noises, such as defective pixels, baseline drifts and random noise, which will greatly degrade the quality of recovered spectra and further reduce the retrieval accuracy of greenhouse gases. In this article, we model the mixed noise removal of interferogram images as an inverse problem and propose a novel optimization framework by integrating the robust principal component analysis (RPCA), unidirectional total variation (UTV), and penalized least squares (PLS) together. Specifically, the low-rankness of interference fringes and sparsity of defective pixels are both considered in the RPCA model, and the directionality of interference fringes is enhanced by the UTV model, while the PLS model guarantees the fidelity and smoothness of the fitting baselines. An efficient alternating direction method of multipliers procedure is designed to solve the proposed framework. A number of experiments were conducted to illustrate the superior performance of the proposed framework in terms of visual quality and quantitative assessment, as compared with the state-of-the-art algorithms.