水下声图像多分辨率融合研究
Underwater Acoustic Image Multi-resolution Fusion Research
-
摘要: 针对水下特殊环境的声图像特点,提出了一种基于梯度金字塔的图像融合算法。首先,对已经配准的源图像进行梯度金字塔分解;然后,在多分辨率下对两幅图像构建基于局部邻域的显著性量测和匹配性量测。在构建显著性量测中,针对高分辨率层采用局部梯度,低分辨率层采用局部方差;在构建匹配性量测中,针对高分辨率层利用局部相关性,低分辨率层利用局部方差比,并且利用图像的局部特征指导融合策略。最后,经过合成模块和多分辨率逆变换得到融合图像。声纳图像融合实验结果证明了所提方法的有效性.Abstract: An image fusion algorithm based on gradient pyramid is proposed which takes advantage of image characteristics of underwater sonar environment .First,the gradient pyramid transform was imposed on registrated sonar images .Second,an activity and match measure were constructed based on the local neighborhood of the images. The local neighborhood feature of images was considered as a fusion strategy. Finally,the fusion image was obtained through synthesis module and multi-scale inverse transform. Experiments revealed that the new algorithm is effective under objective evaluation criteria in the absence of a standard reference image.