【作者】 纪现才；

【导师】 苏开娜；

【作者基本信息】 北京工业大学 ， 计算机应用， 2003， 硕士

【摘要】 视频图像超分辨率复原技术在视频监控、卫星图像等领域具有十分重要的应用价值和广阔的应用前景。当CCD相机在对空间频率较丰富的景物进行采样时，由于CCD像元尺寸的限制，图像分辨率低，混频现象有时很严重。超分辨率图像复原技术可以利用视频图像序列中各帧之间的冗余信息，重构出超分辨率图像，消除和降低混频效应。本文研究和改进了运动补偿迭代算法和共轭梯度最优化迭代算法，从欠采样图像序列中复原出高分辨率的图像，使被重建图像的分辨率比欠采样帧提高了2~4倍。 本文设计了基于多帧输入复原算法的运动补偿迭代算法以提高单色和彩色图像序列的分辨率，并对各种运动估计方法进行了研究和比较。提高视频图像序列的空间分辨率主要取决于运动估计算子的精度，重点放在利用(R,G,B)颜色通道进行运动估计，其主要前提是三个不同的颜色通道传递着相同的运动信息并且能够产生更加精确的运动估计。最后分别对单色和彩色图像序列进行了实验，利用峰值信噪比(PSNR)来度量算法的性能。 在共轭梯度最优化迭代算法中，对相机模型、图像微位移和微旋转角精确配准、共轭梯度重建等关键技术进行了研究。实验结果表明共轭梯度最优化迭代算法是鲁棒的、快速收敛的，并且大量节省内存。此算法易于扩展，使多传感器数据融合成为可能。更多还原

【Abstract】 Super-resolution restoration of video sequences is useful and often critical in many existing applications. Some imaging systems employ detector arrays that are not sufficiently dense to meet the Nyquist criterion during image acquisition. Therefore, the spatial resolution afforded by the optics can’t be fully utilized in such imaging systems. This paper researches and improves two important algorithms to reconstruct high resolution images ,with reduced aliasing, from a sequences of undersampled rotated and shifted frames. The reconstructed image resolution is from two to four times higher than the undersampled frames. In this paper a motion compensation iterative algorithm is designed base on the multiple input restoration algorithm for enhancing the resolution of monochrome and color image sequences. Various approaches toward motion estimation are investigated and compared. Improving the spatial resolution of an image sequences critically depends upon the accuracy of the motion estimator. Particular attention is paid to the use of the color channels in estimating the motion. The main premise is that three different intensity channels conveying the same motion information and yielding more accurate motion estimates. Experiments are performed on monochrome and color image sequences, and performance is measured by the peak signal to noise ratio(PSNR).In conjugate gradient optimization algorithm, the continuous and digital models of an imaging system are defined to explain image acquisition, the image registration algorithm and the conjugate gradient reconstruction algorithm are designed. Experiment results show that the conjugate gradient reconstruction algorithm is robust, rapid convergent, and memory saved. This algorithm is prone to be extended and make it possible that multiple sensor data fusion constructs high resolution images. Ji Xian-Cai[Application of computer] Directed by Su Kai-na更多还原