B超图像最优化重建技术研究  

Research on Optimal Reconstruction Technology of B-ultrasound Image

作　　者：裴锦博 Pei Jinbo(School of Computer and Information Engineering,Henan Normal University,Xinxiang Henan 453007,China)

机构地区：[1]河南师范大学计算机与信息工程学院,河南新乡453007

出　　处：《信息与电脑》2020年第20期17-19,共3页Information & Computer

摘　　要：提高图像分辨率是近年来研究人员关注的热点,尤其在医学图像方面,已有很多研究成果。其中,B超图像在生成以及传输过程中,往往会受到噪声杂音、硬件设备等因素影响导致图像不清晰而影响医学判断。针对此问题,本文依据B超图片的特点利用生成式对抗网络,运用残差块结构,引入了感知损失函数,获取最优的生成图像以完成超分辨率重建。实验结果表明,基于生成式对抗网络的B超图像超分辨率重建能够获得很好的视觉效果,可以达到辅助医生诊断、帮助患者治疗的目的。In recent years,researchers have focused on improving image resolution,especially in medical images.In the process of generation and transmission of B-ultrasound image,it is often affected by noise,hardware equipment and other factors,resulting in unclear image and affecting medical judgment.In order to solve this problem,according to the characteristics of B-ultrasound images,this paper uses the generative countermeasure network,uses the residual block structure,and introduces the perception loss function to obtain the optimal generated image to complete the super-resolution reconstruction.The experimental results show that the superresolution reconstruction of B-ultrasound image based on generative countermeasure network can obtain good visual effect,and can help doctors diagnose and help patients to treat.

关 键 词：B超图像 SRGAN 感知损失函数 残差网络 

分 类 号：R445.1[医药卫生—影像医学与核医学] TP301-4[医药卫生—诊断学]