Curvelet变换域图像融合算法

Algorithm for image fusion in the Curvelet transform domain

本文提出一种基于Curvelet变换域的局部区域能量匹配度的图像融合算法。该算法利用Curvelet变换多尺度和各向异性特征,可在保持图像稳定性的同时增强图像的细节信息,用于同一场景不同聚焦图像的数据融合可获得理想的结果。定量分析了在添加不同方差高斯噪声情况下利用小波和本文方法得到的融合图像对应原标准图像的峰值信噪比(PSNR)。实验结果表明采用本文的图像融合算法在信息熵、结构相似度等方面均好于经典算法。

In this paper, a novel algorithm for image fusion based on the local region energy matching in the Curvelet transform domain is proposed. Utilizing the multiscale and anisotropic property of Curvelet transform, the method can improve the details of an image as well as maintaining its stability. When applied to the images of the same view with different focuses, the method can produce satisfactory result. The peak signal-to-noise ratio is used to compare the new method with reference to the Wavelet method for various levels of Gaussion noise. The experimental results show that the indexes, such as information entropy and structure similar parameter, which are obtained from the Curvelet transform are better than the conventional methods.