二维多光纤光谱图像是一类天文学领域观测获取的目标图像，对这类图像的处理不能依赖于大量的人工干预，而必须借助机器学习、模式识别等技术以高效、自动的方式完成。因此，本项目拟结合机器学习、图像处理领域的智能处理技术，针对二维多光纤光谱图像处理过程中所面临的一些关键问题，深入研究其解决方案。建立新的多光纤光谱成像模型，具有可计算性，得到高效、高质量的优化并求解；发展多源分解算法，将不同源的信号从多光纤光谱图像中分离出来，有效实现噪声消除；建立全新的从图像中提取光谱信号的方法，在进行psf估计的同时同步提取光谱信号。

Multi fiber spectrum image is a kind of objective image which is obtained by telescope in field of astronomy. Processing of these images should not be dependent on massive manual interference, and needs to be done in highly effective and automatic ways using machine learning and model recognition. As a result, this project attempts to deeply study solutions to some key issues in processing of 2D multi fiber spectrum images. By establishing a calculative new 2D fiber spectrum image formation model, we get optimizations with high efficiency and quality thus solutions. We develop a multi source decomposition algorithm to effectively perform noise elimination by separating signals from different sources out of multi fiber spectrum images. We also establish new methods to extract signals from images to extract spectrum signals while doing PSF estimation.