摘要
本文提出了一种高效的、高精度的适合于医学图像增强方法。本研究表明,使用暗特征增强图像会使图像丢失一些细节,加强小尺寸特征会放大噪声,所以仅使用白tophat变换,从递归开运算后图像中提取特征。为解决非对偶运算引起的图像亮度偏差问题,作者提出了条件归一方法。新算法主要特点是形态运算采用非对偶运算,对比度拉伸运算采用加法代替乘法运算。用本算法对MR图像进行了测试,并与已有基于多尺度形态学方法进行比较。实验结果表明,这种算法可以更有效地增强图像的局部对比度,对噪声不敏感,处理后图像精度更高。与以前形态学方法相比,本文方法仅需使用其四分之一特征层就可实现局部对比度增强,从而大大减少了运算量。
The paper presents a more efficient and accurate method for medical image enhancement. In research we find some details can be removed by enhancing dark features and noise be amplified by emphasizing smaller features. So the features are extracted from recursively opened images by white tophat transformation. To avoid some gray-level bias we propose the method of normalized gray-level under condition. In this algorithm, the morphological operations are non-dual and the contrast stretching operations are addition instead of multiplication. The algorithm is tested by MR images and compared with the existing method based on multiscale morphology. The experimental results show that the method is more effective, less sensitive to noise and preserving processed images more accurately.Moreover, the algorithm works with one-fourth features as many as previous morphological method for local contrast enhancement and reduces computational cost.
出处
《北京生物医学工程》
2005年第3期190-194,共5页
Beijing Biomedical Engineering
