基于加权红-黑小波变换的DR图像增强  

Digital radiography(DR) image enhancement based on weighted red-black wavelets transform

作　　者：赵雨[1] 林晓佳[1] 阳维[1] 冯前进[1] 

机构地区：[1]南方医科大学生物医学工程学院,广东广州510515

出　　处：《暨南大学学报（自然科学与医学版）》2015年第3期191-201,共11页Journal of Jinan University(Natural Science & Medicine Edition)

基　　金：国家科技支撑计划项目(2012BAI14B02);国家863计划(2012AA02A616);国家自然科学基金(61471187;31371009)

摘　　要：目的:数字化X线图像(DR)细节常被淹没、动态范围宽.本研究使用了一种基于加权红-黑小波(WRB)变换的增强方法解决这一问题.方法:原始DR图像先进行对数变换,然后对变换后的图像进行WRB分解得到各层的系数;通过设计的分段非线性子带系数操作函数,分别对各层系数进行处理,最后利用WRB反变换和处理过的子带系数重构出增强的图像.结果:WRB方法测试原始DR图像,平均运行时间约为0.6 s;利用该方法对DR图像进行增强后,图像细节显示效果和对比度均得到提升,且无光晕伪影产生;与一些常用的增强算法相比,用WRB算法增强的图像信息熵和交叉熵指标均较优.结论:相比一些传统增强算法,基于加权红-黑小波变换的DR图像增强方法具有明显优势,不仅能有效压缩图像动态范围,还增强图像细节和对比度.Aim：An enhancement algorithm based on weighted red-black wavelets （WRB）transform is presented to solve the problems of details buried and wide dynamic range in original DR images.Meth-ods：logarithmic transformation was first performed on the original DR images.then,the changed images were decomposed within the WRB architecture to get the wavelet coefficients of each layer.Furthermore a piecewise nonlinear coefficient operation function was designed to process the coefficients.In the end, the enhanced images were reconstructed by the inverse WRB transform.Results：The proposed algorithm is tested on the original DR images and compared with some other enhancement algorithms.The average running time of our algorithm is about 0.6 seconds and significantly faster than the other algorithms.U-sing our enhancement algorithm,both of the details and local contrast of the DR images are promoted, meanwhile no halo artifacts have been produced in the enhanced images.Furthermore,the information entropy and cross entropy indices of our algorithm are better than that of the other algorithms.Conclu- 〈br〉 sion：The proposed enhancement algorithm is an effective and efficient method for compressing image dy-namic range,enhancing image details of the original DR images.

关 键 词：DR图像 图像增强 加权红-黑小波(WRB) 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]