基于多波段匹配的超光谱成像仪图像条带噪声去除方法研究

Destriping of Hyperspectral Image Based on Multi-band Image Matching

环境卫星超光谱成像仪前20个通道的图像具有明显的条带噪声,严重影响了后续数据处理和定量化应用。根据环境卫星超光谱成像仪波段多、各波段图像相关性强的特点,提出了一种基于多波段图像匹配的条带去除方法。以广西北部湾地区的超光谱成像仪图像为例,分别利用矩匹配方法和多波段匹配方法进行校正。最后从图像基本信息保留能力、图像灰度值改变量和列均值比评价指标三个方面,定量评价了这两种方法的去噪效果。结果表明,多波段匹配方法能够在有效去除图像条带噪声的同时,较好地保留各探测元接收到的地物信号的差异,其去噪效果优于矩匹配方法。

There are obvious striping noises in the images obtained by the first twenty channels of HJ-1A satellite.These striping noises have great influence on the subsequent data processing and quantitative application of the images.A destriping method based on multi-band image matching is proposed according to the fact that the hyperspectral imager onboard the HJ-IA satellite has many bands and the image in each band has great correlation.An image of Beibu Gulf region in Guangxi is corrected by using the moment matching method and the multi-band matching method respectively. Finally,the destriping effectiveness of those two methods are evaluated quantatively in the aspects of basic image information retaining,DN value change and line mean value ratio.The result shows that the the multi-band matching method can remove the striping noises in images effectively while retaining the difference of the signals in each detection element.It is better than the moment matching method.