基于小波系数相关性的MRI图像像素点分类处理算法  

A Pixel Classification Method in MRI Enhancement Based on Wavelet Coefficient

作　　者：程科[1] 汪正霞[1] 孙玮[1] 

机构地区：[1]江苏科技大学计算机科学与工程学院,江苏镇江212003

出　　处：《南京师大学报（自然科学版）》2014年第1期41-46,共6页Journal of Nanjing Normal University(Natural Science Edition)

基　　金：国家自然科学基金(51008143);江苏省科技创新与成果转化(重大科技成果转化)项目(BA2012129);江苏省研究生科研创新计划资助项目(1252209AK)

摘　　要：由于图像软组织的对比度和信噪比均较低的原因,传统的二进小波算法在进行磁共振图像的增强处理时效果不明显.文章首先在分析小波变换对噪声影响规律的基础上,结合磁共振图像(MRI)特点,对二进小波变换中各尺度上小波系数间的相关性进行分析和归类.其次,文章通过改进的小波系数相关性置信度方法,将MRI图像的像素点分为信号点、噪声点和性质未定点,并采用不同的增强函数对信号点和噪声点进行有效处理.最后,文章提出了一个基于三阶分段函数的性质未定点处理算法.实验表明,文章中所采取的方法能较好地实现像素点的分类,抑制其中的噪声点,有效地增强目标像素点.Dyadic wavelet transforms have no demonstrable effect on MRI enhancement,because MRI has the disadvantage of its lower contrast and noise-signal ratio. After analyzing the effects of wavelet on signal noise and based on the characteristics of MRI,this paper analyzes the correlation of dyadic wavelet transform of wavelet coefficient in each scale and classes them to different types. Then a revised method of wavelet coefficient confidence is introduced,hence MRI pixels are classified to three types, including singal itself, noise and undefined pot. At last, after using different functions on enhance steps for the first two types of signals,a new piecewise function has been raised to deal with the undefined pot. Experiment shows that using this method can effectively classify pixels,reduce the noise and enhance the true signal pixel.

关 键 词：二进小波 磁共振图像 图像增强 多尺度积 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]