摘要
讨论了图像成像的基本模型,并提出了一种基于调整核回归函数作为正则项的序列图像重建算法。该算法是对已经提出的核回归算法的改进,减少其在超分辨率图像重建时的运算量。而且在图像配准过程中针对图像间只存在平移和旋转变换,采用了基于矩形像素值的亚像素配准方法,以提高配准的速度和精度。利用此算法对序列图像进行重建仿真,并通过结论得出其在噪声严重的情况下具有更好的边缘保留特性。
The basic imaging model is discussed, and a novel algorithm based on steering kernel regression method as a regularization term for super resolution reconstruction of sequence image is proposed. This algorithm is improved from the kernel regression method which has proposed and reduces calculation cost when it used in image super resolution construction. During the image registration, the algorithm that bases on the rectangular surface samples pixel of image sub-pixel registration is used to improve the process rate and precision. In the end, experimental results demonstrate that the effectiveness of our method has a better edge-preserving under high noise effect.
出处
《无线电工程》
2009年第4期17-19,共3页
Radio Engineering
关键词
图像重建
调整核回归
双边滤波器
亚像素配准
正则化项
image reconstruction
steering kernel regression
bilateral filter
sub-pixel registration
regularization term
