为了有效地克服单波段前视红外图像中存在的点状杂波、 条状波浪以及局部高亮区域等随机杂乱背景的影响, 开展了基于多波段前视红外图像融合的海面杂乱背景平滑方法的研究。 充分利用多波段前视红外图像之间的互补性和差异性, 通过融合多波段红外图像的信息, 旨在平滑抑制海面杂乱背景并保持舰船目标的特征信息, 为舰船目标检测提供一幅优质的图像。 首先利用离散小波变换将多波段源图像分解为低频子带和高频子带, 其中, 高频子带主要包含了图像中背景以及舰船目标的细节信息, 低频子带主要包含了图像的亮度以及对比度信息; 对于高频子带, 在基于高频系数取绝对值最大法得到高频融合图像后, 计算每个像素的区域能量来对高频融合图像进行调制以抑制图像背景的细节信息而保留舰船目标的细节信息; 对于低频子带, 通过平均法融合低频子带并利用导向滤波对低频融合图像进行平滑滤波处理; 最后对高频融合图像和低频融合图像进行小波逆变换得到的重构图像即为融合图像。 对实际采集的多波段前视红外图像进行仿真实验, 将该方法与双边滤波、 导向滤波、 梯度最小化、 相对全变分、 双边纹理滤波和滚动滤波共6种图像平滑滤波方法进行对比。 结果表明: 所提出的方法通过有效地融合多波段图像的信息, 将空间域的平滑处理转换到频率域中进行, 能够很好地平滑海面随机杂乱背景并较好地保持舰船目标的结构、 灰度以及对比度信息, 大大增强了舰船目标的可分离性, 其图像平滑性能优于作为对比的6种方法。

For effectively overcoming the influence of rough background including point-clutter, strip wave and highlighted area in single-spectral forward-looking infrared (FLIR) image, a smoothing method for sea surface rough background based on multi-spectral FLIR images fusion is proposed. The method makes full use of the complementarity and difference existing in multi-spectral FLIR images. It aims at combining multiple images into a quality image in which the sea surface rough background is smoothed and the feature information of the ship targets is maintained good. Firstly, the multi-spectral source images were decomposed into low frequency sub-bands and high frequency sub-bands by discrete wavelet transform (DWT). The high frequency sub-band mainly contains the detailed information of the background and the ship target while the low frequency sub-band mainly contains the grayscale information. After obtaining the high frequency fusion image based on the maximum value of the high frequency coefficient, the regional energy of each pixel was calculated to modulate the high frequency fusion image in order to suppress the details of the background and maintain the details of the ship targets simultaneously. Then the low frequency fusion image was obtained by the average strategy and smoothed by the guided filter. Finally, the fusion image was reconstructed based on the high frequency fusion image and the low frequency fusion images by inverse wavelet transform. When the simulation experiment was carried out on the actually collected multi-spectral FLIR images to prove the effectiveness of the proposed method, the proposed method was compared with the other 6 smoothing methods including bilateral filter, guided filter, gradient minimization, relative total variation, bilateral texture filtering and rolling guidance filtering. A large number of experimental results show that the smoothing performance of the proposed method is better than the other 6 methods. The proposed method can effectively smooth the sea surface rough background and maintain the structure, grayscale, contrast of the ship targets, which can greatly enhance the separability of the ship targets. In the future work, the proposed method needs to be optimized to further improve the timeliness.