摘要
研究图像配准精确度优化提高问题。图像配准技术一直被广泛应用在医学图像和遥感图像领域中。但由于同一目标不同信息来源的图像之间存在差异,配准造成图像不清晰。传统的图像配准算法效率和精度较低,特别是传统算法的计算复杂度高。为了解决上述问题,提出了一种将改进的曲线傅里叶变换图像配准算法,有效结合了最大熵算法和傅里叶变换算法,采用傅里叶变换算法对图像中感兴趣的区域进行分割,对各个分割区域特点进行描述并组成一定的结构,然后用最大熵算法进行权值训练,从而得到精准的图像配准结果。仿真结果表明,改进的算法有效的提高了图像配准的精确度,验证了改进算法是一种可行性有效的图像配准方法。
Research image registration accuracy optimization.Improved image registration technology has been widely applied in medical image and remote sensing image.The traditional image registration algorithm is of low efficiency,low precision and high complexity.To solve the above problem,this paper proposed an image registration algorithm based on improved Fourier transform curves,which efficiently combined maximum entropy and Fourier transform algorithmd.First,Fourier transform algorithm was used to segment the interest region of image,and the characteristics of each segment were described and composed to certain structure.Then,the maximum entropy algorithm was used to train weithts so as to obtain precise image registration results.Simulation results showed that the improved algorithm can improve the image registration accuracy,and verify the feasibility of the image registration method.
出处
《计算机仿真》
CSCD
北大核心
2011年第10期265-268,共4页
Computer Simulation
关键词
傅里叶变换
图像配准
最大熵算法
权值训练
Fourier transform
Image registration
Maximum entropy algorithm
Weitht training