自适应正则化多幅影像超分辨率重建

Adaptively regularized muti-frame image super-resolution reconstruction

影像超分辨率技术已经成为近年来影像处理领域的研究热点。其中,正则化重建模型由于具有求解模型直观、解唯一等优点而得到了广泛应用。在正则化重建模型求解过程中,正则化参数对于重建结果的好坏有着重要影响,参数选择过小就不能很好地抑制噪声,参数选择过大又会模糊重建影像。将数值计算领域的U曲线方法引入到超分辨率重建领域,用来确定重建模型中的最优正则化参数。首先建立U曲线,然后选择U曲线的左侧曲率最大点所对应正则化参数为重建正则化参数。实验结果表明,无论是在目视效果还是定量评价方面,重建结果都优于传统的自适应迭代方法和L曲线方法。

Image super-resolution reconstruction has been a hot research topic in recent years. Among kinds of reconstruction methods, regularized reconstruction is widely used, because it applies simple principle and unique solution. The regularization parameter plays an important role in reconstruction. If the parameter is too small, the noise will not be effectively restrained, conversely, the reconstruction result will become blurry. Therefore, a U-curve based reconstruction method is proposed, using the unique features of U-curve to select the regularization parameter. The data fidelity term and a prior item are used to form a U-curve function, and the left maximum curvature point is selected as the optimal regularization parameter. The proposed method is tested on two simulate data sets. The results show advantages of this revised method both in visual effects and quantitative evaluation.