白天观星红外星图噪声抑制算法

Denoising algorithms to infrared star map of daytime star observation

单点噪声和条纹非均匀性噪声是白天观星红外星图中的两种典型噪声,严重影响了星图识别的后续处理。针对两种噪声产生的原因和特点,提出改进的单点噪声去噪算法,即用局部窗口遍历扫描全图检测出单点噪声,然后用全局均值进行补偿;又提出列直方图偏移校正算法去除条纹非均匀性噪声,即依据列偏移校正模型,对每列直方图相对于全局直方图的峰值位置偏移进行校正。实验结果表明:所提出的噪声抑制算法对红外星图去噪效果良好,星点目标峰值信噪比可从7.4提高到12.8,并且算法计算简单,可以在单帧时间内完成,满足星图实时预处理的应用要求。

Single-point noise （SPN） and stripe non-uniformity （SNU） noise are two typical noises in infrared star map of daytime star observation, which seriously affect subsequent processing of star recognition. Considering the causes and characteristics of the two noises, an improved SPN denoising algorithm was proposed to denoise SPN. The full map was scanned by a local window to detect SPN, and the SPN was instead by global mean value. And a colum histogram offset correction （CHOC） algorithm was proposed to denoise SNU noise. According to the colum offset correction model, the offset of histogram peak position between colum histogram and global histogram was corrected. Experimental results indicate that the algorithms proposed could achieve good denoising effects with infrared star map, and the peak signal-to-noise ratio of star point target could be raised from 7.4 to 12.8, meanwhile, the algorithms could finish in one frame time with low computation complexity, so the algorithms meet the requirements of real-time star map preprocessing application.