基于交互式数据语言MRI图像直方图增强方法的比较  

Comparison of histogram enhancement approaches to MRI image based on interactive data language

作　　者：王娟[1] 王胜菊[1] 汤乐民[1] 

机构地区：[1]南通大学医学院影像工程教研室,江苏省南通市226001

出　　处：《中国组织工程研究与临床康复》2008年第44期8657-8660,共4页Journal of Clinical Rehabilitative Tissue Engineering Research

基　　金：南通大学自然科学研究计划项目(06Z088)~~

摘　　要：目的:比较不同直方图增强方法对改善MRI图像质量的作用效果。方法:基于交互式数据语言编程环境,分别采用直方图均衡化、自适应直方图均衡化和直方图规定化对颈髓MRIT2W图像进行增强处理,利用峰值信号噪声比和图像信息熵评价图像暗区细节的表现性能以及图像噪声水平。结果:直方图均衡化对图像暗区细节的增强效果一般,图像整体对比度反而有所下降;自适应直方图均衡化增强细节的同时放大了图像噪声,并在边缘形成伪影;直方图规定化可选择匹配直方图函数的类型,充分显示图像暗区细节,图像噪声水平低于前两种方法。结论:应用直方图增强方法处理颈髓MRIT2加权图像,在图像暗区细节表现和图像低噪声水平上,直方图规定化明显优于直方图均衡化和自适应直方图均衡化。AIM： To compare the effects of different histogram enhancement algorithms on improving the quality of MRI image. METHODS： Three processing algorithms, including histogram equalization, adaptive histogram equalization and histogram specification, were applied to enhance a MRI cervical spine T2-weighted image based on programming software interactive data language. The capability of representation of details in dark area and the level of noise were evaluated by means of peak signal to noise ratio and image information entropy. RESULTS： Histogram equalization cannot enhance the details in dark region obviously, but decline the contrast of whole image; adaptive histogram equalization can improve details but enlarge noise and engender shadow at edges simultaneously; histogram specification can choose the type of histogram function to match; it reveals the details in dark area sufficiently, and there is the lowest level of noise among these three algorithms. CONCLUSION： MRI cervical spine T2-weighted image processing with different algorithms of histogram enhancement, histogram specification is more outstanding than histogram equalization and adaptive histogram equalization in the representation of details and the low-level of noise.

关 键 词：直方图均衡化 自适应直方图均衡化 直方图规定化 交互式数据语言 

分 类 号：R318[医药卫生—生物医学工程]