基于Canny算子加权引导滤波的Retinex医学图像增强算法  

Retinex Medical Image Enhancement Algorithm Based on Weighted Guided Filtering by Canny Operator

作　　者：陈韵竹 郭剑辉[1] CHEN Yunzhu;GUO Jianhui(School of Computer Science and Engineering,Nanjing University of Science and Technology,Nanjing 210094)

机构地区：[1]南京理工大学计算机科学与工程学院,南京210094

出　　处：《计算机与数字工程》2019年第2期407-411,480,共6页Computer & Digital Engineering

基　　金：国家自然科学基金项目(编号:61603190);江苏省自然科学基金青年基金项目(编号:BK20140794)资助

摘　　要：医学图像处理和分析在对疾病的成像诊断中起到辅助医生确定病变局域的重要作用。但是由于人体的组织密度极其复杂,由于各种不良因素的影响,从而导致医学图像(尤其是X射线医学影像等)普遍存在对比度低、动态范围窄、强度分布不对称、边缘不清晰等问题。针对上述问题,结合医学图像特点和光照估计模型提出一种基于Canny算子加权引导滤波的Retinex(CWGFR)医学图像增强算法。首先,利用Canny边缘检测算子准确估计加权引导滤波器的边缘权重并通过加权引导滤波器对光照进行估计,从而得到入射光分量;然后根据Retinex算法原理计算出反射光分量,即log域上的增强图像;最后对图像进行量化处理,增强其图像对比度从而得到输出图像。对比实验结果表明,论文算法在提高图像对比度的同时,具有更高的边缘保持特性,能够凸显图像细节信息,有效消除"光晕"伪影。Medical image processing and analysis plays an important role in assisting the physician in determining the patholo-gy in the imaging diagnosis of the disease. However,due to the extremely complex tissue density of the human body and various ad-verse-factors,medical images(especially X-ray medical images,etc.)generally have problems such as low contrast,narrow dynam-ic range,asymmetric intenstity distribution and unclear edges. Aiming at the above problems,a Retinex medical image enhance-ment algorithm(CWGFR)based on weighted guided filtering is proposed combined with medical image feature and light estimationmodel. Firstly,the canny operator is used to estimate the edge weight of the weighted guided filter accurately. Secondly,the light isestimated by the weighted guidance filter to obtain the incident light component. Then,the reflected light component(enhanced im-age in log domain)is calculated according to the Retinex algorithm. Finally,the image quantization processing enhances its imagecontrast to get the output image. The experimental results show that the proposed algorithm can improve the image contrast. In ad-dtion,it also has higher edge preserving characteristics,and can preserve and highlight the details of the image.

关 键 词：医学图像增强 RETINEX 加权引导滤波 边缘感知 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]