基于全变分模型和高斯曲率滤波的红外图像条纹噪声去除算法

An infrared image stripe noise removal method based on total variational model and Gauss curvature filtering

针对红外热像仪读出电路的偏置电压存在非均匀性,造成红外图像出现条纹噪声的问题,提出了基于全变分模型和高斯曲率滤波结合的去噪算法。在分析红外条纹噪声成因并研究其特性的基础上,首先对含噪图像采用全变分模型进行去噪处理;然后确定复原图像出现阶梯效应的区域,将其对应的噪声图像中的区域看作可展曲面,采用高斯曲率滤波处理;最后将全变分模型和高斯曲率滤波的处理结果综合输出。实验结果表明,所提算法能够去除红外图像中的条纹噪声,并且能够克服全变分去噪后复原图像出现阶梯效应的问题。

Aiming at the non-uniformity of bias voltage in the readout circuit of infrared thermal imager,which re- sults in stripe noise in infrared image,a denoising algorithm based on total variational model and Gauss curvature filte- ring is proposed in this paper. On the basis of analyzing the causes of infrared stripe noise and studying its characteris- tics,the total variational model is used to denoise the noisy image. Then the region where the step effect appears in the reconstructed image is determined,and the region in the corresponding noise image is regarded as an expandable sur- face,which is processed by Gauss curvature filtering. Finally,the processing results of the total variational model and Gauss curvature filtering are synthetically outputted. Experimental results show that the proposed algorithm can remove stripe noise in the infrared image and overcome the step effect of the reconstructed image after total variation denoising.