摘要
针对电容层析成像系统重构图像问题的非线性、病态性等特点,提出了基于径向基神经网络的图像重构算法以达到流型辨识的要求,并针对神经网络成像精度低的问题,重点分析研究了适用于图像特征增强的空频分析方法——基于径向基网络的自适应小波滤波重构算法.图像经过多级小波分解后,依据邻域特性来判断各个子带图像上的边缘点;依据统计特性来估计噪声在空频域中的分布,并构造具有自适应性的边缘增强增益;最后用调整后的小波系数重建图像.仿真结果表明,该算法有效地增强了图像的边缘,减小了图像的噪声影响,很大程度上改善了重构图像的成像质量,明显减轻了失真度,使图像特点更加清晰.
For nonlinearity and morbid state of image reconstruction for electrical capacitance tomography system, a radial basis function network image reconstruction algorithm is proposed to fulfill the requirement of flow regime identification. And the adaptive wavelet filter image reconstruction algorithm based on RBF, which belongs to a space-frequency analysis method suitable for image feature-enhanced, is emphatically analyzed to heighten the reconstruction accuracy. Undergoing multi-level wavelet decomposition of the image, the edge points are determined based on the neighborhood properties of each sub-band, the noise distribution in the space-frequency domain is estimated following the statistical characteristics and a self-adaptive edge enhancement gain can be obtained to reconstruct the image with the adjusting wavelet coefficients. The simulation results demonstrate that this algorithm enables to effectively implement image enhancement and edge detection, greatly improve the quality of reconstructed image and obviously reduce the distortion.
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
2010年第2期1-5,共5页
Journal of Xi'an Jiaotong University
基金
国家自然科学基金资助项目(50777049)
国家"863计划"资助项目(2009AA04Z130)
国家"973计划"资助项目(2005CB2212206)
关键词
图像重构
径向基网络
小波滤波
流型辨识
image reconstruction
radial basis function network
wavelet filter
flow regime identification
