【摘要】 基于压缩感知理论提出了一种红外与可见光图像的融合新方法。该方法将Contourlet变换(CT)和小波变换(WT)相结合,以进一步增加变换后系数的稀疏性,同时对采样模式和融合规则进行改进。首先对图像进行Contourlet变换,再对各高层分解系数进行正交小波变换;然后使用各层采样率不同的分立双放射形采样矩阵对系数采样,并用不同的规则对各层采样值进行融合;最后使用非线性共轭梯度法重构融合图像。实验结果表明,在采样率为0.5时,本文方法融合图像的细节信息比小波方法和小波变换压缩感知(WTCS)方法更加丰富;在所有采样率上,本文方法的融合效果比WTCS法在互信息、空间频率和融合信息逼真度等客观融合质量评价指标上均提高约10%。更多还原

【Abstract】 A novel method to fuse infrared and visible images was proposed based on compressive sensing theory.The method combined Contourlet Transform(CT)with Wavelet Transform(WT)to increase the sparsity of transformed coefficients and also to improve sample patterns and fusion rules.Firstly,the original images were decomposed in a Contourlet domain,and orthogonal wavelet transform was applied to the high level decomposed coefficients.Then,the composite double radially sampling mode with different sampling rates in each decomposition level was used to perform the linear measurements of coefficients and to fuse the measurement values using different rules in eachlevel.Finally,the fused image was reconstructed by using nonlinear conjugate-gradient solution.The experimental results demonstrate that the detail information of fusion image by proposed method is more salient than that of discrete wavelet transform fusion image when sampling rate is 0.5.As compared with WTCS method,the mutual information,spatial frequency and the visual information fidelity of fused image from proposed method are increased by 10%.更多还原