采用变分法的遥感影像条带噪声去除

Stripe Noise Removal of Remote Images Based on Variation

为了避免去除遥感影像条带噪声时丢失影像细节,构建了综合两种单向变分模型的组合去噪方法。首先,构建一种条带去除单向变分模型,通过此模型可以有效去除遥感影像条带噪声,从而初步获得近似遥感影像及条带噪声影像,同时避免近似遥感影像出现阶梯效应;其次,为了避免条带噪声影像中包含的少量细节信息丢失,构建一种条带保留单向变分模型,通过此模型可以去除条带噪声影像中的细节信息而只保留条带噪声,从而有效分离出细节信息;最后,把分离出的细节信息与近似遥感影像进行累加得到去噪影像。通过实验分析发现,本文方法不仅能完全去除条带噪声,而且基本没有破坏影像细节,去噪后的影像质量得到明显提高。

To avoid the loss of image details when removing the stripe noise of remote images, a combined destriping method of two unidirectional variation models was constructed. First, a stripe removal unidirectional variation model which can effectively remove the stripe noise and obtain the approximate restoration images and stripe noise images was constructed, and the stair effect is avoided in approximate remote images. Second, in order to avoid losing details in the stripe noise image, a stripe preserve unidirectional variation model was constructed, which could remove the image details from the stripe noise image and only preserve the stripe noise, so that the stripe noise and the image details could be separated effectively. Finally, the approximate remote image and the image details were superimposed to obtain the destriped image. Experimental results show that the proposed method can effectively restrain the stripe noise of remote images, and preserve the details of the remote images very well, hence the quality of destriped images is improved obviously.