摘要
研究一种能够有效控制图像增强最佳尺度选取的方法对改善单幅医疗图像后续的目标检测和识别具有重要意义。针对现有方法由于无法将医疗图像中的目标信号保留的同时抑制背景噪声,导致增强效果不佳,且无法实现图像增强最佳尺度选取精度控制等问题,提出一种基于非线性处理的单幅医疗图像增强最佳尺度选取控制方法,对医疗图像进行颜色空间转换并划分为若干个图像块,对各个图像块进行结构元匹配获得映射子图,计算映射子图中的结构元特征向量、连通粒属性以及连通粒属性特征,通过对三个特征分量的归一化处理和加权合并,实现单幅医疗图像融合特征提取。采用软阈值函数对医疗图像高频子带部分进行平滑滤波,设置非线性函数阈值对提取的融合特征向量中幅值大于设置阈值的系数进行增强,对于小于设置阈值的系数进行抑制,并通过对亮度系数的调节和非线性函数阈值的选取进行医疗图像增强最佳尺度选取精度控制。仿真结果表明,该上述方法能够实现医疗图像增强最佳尺度选取精度控制,且具有比其他方法更好的增强效果。
In this article,we focus on a method to control optimal scale selection of single medical image enhancement based on non-linear processing method. At first,we transformed color space of medical image and divided it into several image blocks. Then,we matched structural element of each image block to obtain the mapping sub-graph.Meanwhile,the structural element feature vector,the connected particle attribute and the connected particle attribute feature in matching sub-graph were calculated. Through the normalization and weighted combination of the three feature components,we achieved the fusion feature extraction of single medical image. Moreover,we used soft threshold function for the smoothing filtering of high-frequency sub-bands in medical image and set the nonlinear function threshold to enhance the coefficient whose amplitude was larger than the setting threshold in the extracted fusion feature vectors,and to suppress the coefficient which was smaller than the setting threshold. Through the adjustment of brightness coefficient and the selection of nonlinear function threshold,we finally completed the accuracy control of optimum scale selection of medical image enhancement. Simulation results show that the proposed method can achieve the accuracy control of optimum scale selection of medical image enhancement. Meanwhile,this method has better enhancement effect than other methods.
作者
罗梦迪
LUO Meng-di(Medical College,Xuehang University,Xuchang Henan 461000,China)
出处
《计算机仿真》
北大核心
2019年第1期259-262,283,共5页
Computer Simulation
关键词
单幅
医疗图像
增强
最佳尺度选取
控制
Single frame
Medical image
Enhancement
Optimum scale selection
Control
