论文部分内容阅读
针对合成孔径雷达(SAR)图像和可见光图像融合问题,提出一种基于非下采样剪切波变换域的隐马尔可夫树模型的图像融合方法 (NHMM),图像经过非下采样剪切波变换(NSST)分解形成一个低频子带和多个高频子带.在NSST域中,对低频系数采用基于标准差的融合策略;针对高频子带,建立NSST域隐马尔可夫树(HMT)模型对高频系数进行训练,并根据梯度能量对训练后的高频系数进行选择,最后通过NSST逆变换得到融合图像.实验结果表明,所提出的方法可提高图像的融合质量,并能降低图像噪声,具有一定的有效性和实用性.
Aiming at the problem of synthetic aperture radar (SAR) image and visible light image fusion, an image fusion method based on non-subsampling shear wave transform domain (HMMM) is proposed. The image is transformed by non-subsampling shear wave transform (NSST) to form a low frequency subband and a number of high frequency subbands.In the NSST domain, a standard deviation based fusion strategy is used for low frequency coefficients. For the high frequency subband, an NSST Hidden Markov Tree (HMT) The model trained the high frequency coefficients and selected the high frequency coefficients after training according to the gradient energy.At last, the NSST inverse transform was used to get the fusion image.The experimental results show that the proposed method can improve the fusion quality and reduce the image Noise, with a certain degree of effectiveness and practicality.