【摘要】 提出了一种具有平移不变性的图像融合方法,并将其用于战场打击效果评估.算法引入具有平移不变性和良好的方向分析能力的双树复小波变换对源图像进行多尺度分解;然后对分解得到的各尺度的高频子带采用基于跨尺度的邻域空间频率的融合策略,最高层的低频子带采用基于改进的邻域熵的融合策略进行融合处理;最后对融合后的各个频带进行双树复小波逆变换,重构出融合图像.打击效果实验和评估结果表明该算法能够实现更有效的融合,融合图像包含了打击前后的有用信息、标记了打击的位置,更容易对打击的位置或关键部位,以及毁伤程度进行评估.更多还原

【Abstract】 An invariant image fusion technique was presented and applied in battle damage assessment (BDA). Shift invariant and directionally selective dual tree complex wavelet transform was introduced for source image′s decomposition. Two fusion measurements, across-band-neighborhood-space-frequency and improved-neighborhood-entropy were defined for highpass and lowpass subbands respectively. The fused image was obtained by taking inverse dual tree complex wavelet transform. BDA experimental and evaluation results demonstrated that satisfactory results are obtained using the proposed method. The fused image contains useful information about source images and marks the damage positions and key parts, which is valuable for BDA application.更多还原