基于低秩三分解的红外图像杂波抑制

针对红外图像中对比度较低、目标信号较弱且受背景噪声杂波干扰较大的特点,结合信号的稀疏表示理论提出了一种基于低秩三分解模型的红外图像背景杂波抑制算法。首先,分别对红外图像中目标、背景和噪声3种成份进行建模描述,得到低秩三分解模型。然后,采用二维高斯模型构造红外小目标超完备字典,利用所提出的低秩三分解模型将分块重置的图像数据矩阵分解为背景、噪声和目标3种成份。最后,对于目标分量进行阈值处理从而得到突出红外小目标的重构图像,实现杂波抑制。在3种不同情况下的实验结果表明:本文算法能够使红外图像局部信噪比提高2倍以上;与其他经典算法相比,抑制因子至少提高15%。得到的结果表明,所提算法能够有效抑制杂波,在提高红外图像信噪比的同时,对不同噪声干扰也具有较强的鲁棒性。

Clutter suppression of infrared image based on three-component low-rank matrix decomposition

To solve the infrared target detection problems caused by low contrast, weak target signals and background clutter interference, a clutter suppression method based on three-component low-rank matrix decomposition model was proposed combined with the sparse representation theory. Firstly, the three components, including target, background and noise, in a infrared image were described respectively to obtain the three-component low-rank matrix decomposition model. Then, an over-complete dictionary for modeling a small target was constructed by using two-dimensional Gaussian model. The three-component low-rank matrix decomposition model was used to decompose the block reset image data into the background, noise and target components. Finally, the target component was processed by thresholding to obtain a reconstructed image with protruded infrared targets and to complete the clutter suppression. The experiments under three conditions demonstrate that the proposed method has increased the local signal to noise ratio of image more than 2 times, and the background suppression factor has increased more than 15% as compared with that of the classical methods. It con cludes that the proposed method not only suppresses the background clutter, improves the signal-to-noise ratio of the infrared image effectively but also has strong robustness against the noise interference.