在使用传统的滤波去噪方法对红外图像进行预处理时, 会在削弱噪声信息的同时模糊了图像中目标和背景的边缘成分, 不利于对图像信息进一步处理。针对该问题, 提出基于图像融合的红外图像预处理方法, 通过对同一幅退化降质的红外图像分别实施改进的中值滤波算法和改进的 Sobel算子边缘提取算法, 进行去噪处理和边缘提取, 最后利用图像融合算法, 将不同算法处理的结果图像加以合并, 形成融合后图像。实验结果表明, 所提出的基于图像融合的预处理方法能够滤除红外图像中大部分噪声, 且很好地保留了目标和背景的细节特征, 而且运算简单, 便于硬件实时计算, 有利于后续的目标检测与识别处理。

In the traditional filtering method to preprocess the image, the edges of target and background are usually blurred while the noise information being weakened, which is not conducive to the further processing of the image information. Aiming at this problem, this paper proposes image preprocessing techniques based on image fusion: firstly applying improved median filtering algorithm and improved Sobel edge extraction algorithm respectively on the same degraded infrared image, then using the pixel level image fusion algorithm to combine the two processing results. The final fusion results show a clear image with significantly enhanced edge portion with a simple operation. In a word, this preprocessing method is conducive to the target detection and recognition in following treatment and achieves the image preprocessing requirements.