摘要
在超分辨图像复原处理中,像素内的配准精度、速度是超分辨图像最终实现的实时性和高质量的关键因素。传统的利用泰勒级数展开像素内配准方法实时性较差;分级少的块匹配配准精度相对较低。该文提出一种基于分段双三次多项式拟合的分级块匹配像素内配准技术,在算法复杂度低的同时保证了配准精度,并在文中给出了实验统计结果。
In the research of super-resolution image processing, the precision and speed of the sub-pixel registration are the key factors. It is difficult for the Taylor's series expansion method to do in real-time, and the hierarchical block-match method with less levels can achieve only a lower precision. The bicubic curve function method with hierarchical block-match drawn in this paper can not only obtain a higher precision, but also be implemented in real-time.
出处
《电子与信息学报》
EI
CSCD
北大核心
2007年第1期47-49,共3页
Journal of Electronics & Information Technology
关键词
图像处理
超分辨
像素内
配准
Image processing
Super-resolution
Sub-pixel
Registration