为了提高图像分辨率，从软件的角度出发，对低分辨率图像序列重建高分辨率图像的原理和算法实现开展了研究，提出从多个低分辨率图像序列中获取更高分辨率图像的方法。通过采用基于光流场的金字塔分层结构实现由粗到精的图像配准，获取了亚像素的运动估计。在采用多帧低分辨率图像进行亚像素级配准后，提出采用动态自适应确定正则化参数的方法，构造了简单的正则化代价函数，建立了低分辨率图像与高分辨率图像之间关系的重建模型，仿真实验结果表明该超分辨率重建算法水平和垂直方向上分辨率各增加一倍，与其他算法相比清晰度更高，计算时间不到传统POCS算法的一半。

In order to improve image resolution by software processing, this paper investigates the theory and algorithms for reconstructing high resolution images from low resolution image sequences, and proposes a method to obtain high resolution images from low resolution image sequences. We realized the rough-to-fine image registration by pyramid multi-layer construction based on optical flow field, and obtained the sub-pixel motion estimation. After the multi-frame low resolution image was sub-pixel registered, we proposed a method for dynamically and adaptively determining the regularization parameter, constructed the simple regularization cost function and built the reconstruction model between low resolution image and high resolution image. The experimental results show that the horizontal and vertical resolution achieved with this algorithm is two times better than that of the original image. Compared with the traditional POCS algorithm, this algorithm reconstructs clearer image, and the computation time is reduced by 50%.