非下采样剪切波的红外偏振图像融合

针对传统红外目标探测技术中的探测率较低,无法分辨红外伪装的军事目标等缺点,提出区域显著性的非下采样Shearlets变换(NSST)方法融合图像,运用NSST能够将图像精细的分解在频率域,在频率域根据图像显著性分布权重确定融合规则,经过NSST反变换得到融合后图像.结合中波与长波红外实拍实验,比较该方法与其他常用融合方法,结果表明,区域显著性的NSST融合图像噪声信息少、对比度、信息熵及互信息值均较高,方法适合偏振图像的融合,有利于目标的识别与探测.

Infrared polarization image fusion with non-sampling Shearlets

Due to the low contrast, lack of details and difficulties to distinguish target from background in traditional infrared(IR) imaging systems, the detection and recognition probability of camouflage infrared target is relatively low. The polarization image represents the polarization properties of the target. By the fusion of infrared polarization image and infrared image, the infrared target can be enhanced effectively and have a better visual character. In this paper, an infrared image fusion method using non-sampling Shearlets transform (NSST) and regional saliency analysis was proposed. Firstly, the infrared image was decomposed into several sub frequency bands by NSST. Then in the frequency domain the fusion rules were determined by regional saliency analysis. At last, final fused image was obtained by inverse NSST. Middle wave and long wave infrared experiments were done to verify the effectiveness of the presented method. Compared with other image fusion algorithms, both the subjective and objective evaluation results illustrate that the proposed approach could obtain better infrared fused images with less noise, higher contrast and larger information entropy and more mutual information. It has a good application in detection and recognition for infrared targets.