摘要
电子助视系统是一种运用实时图像处理技术来提高低视力患者阅读能力的仪器;由于传统电子助视系统使用单摄像头采集图像,视角较小,要多次移动助视设备或目标书籍才能完成阅读,因此用户体验不佳;针对此问题,设计了一种基于多路摄像头采集的文档图像拼接算法;在图像配准步骤,提出了一种针对文档图像的区域特征描述方法,提取单个文字区域作为特征区域,其质心作为特征点,通过描述特征区域的纹理信息完成特征点的描述,提高了特征匹配效率;在图像融合步骤,提出了一种利用文档图像字符间隙的最优拼接缝搜索算法,消除了拼接鬼影;构建多尺度的图像金字塔,使用多分辨率融合算法,抑制了融合图像的曝光差异;从实验结果来看,算法获得了很好地配准和融合效果,提升了配准速度,消除了配准鬼影和曝光差异,能够获得较广的阅读视角,可以较好地用于电子助视系统。
Electronic video magnify system is an instrument, using real--time image processing technology to improve the ability of reading of low vision people. However, traditional electronic video magnify system usually uses a single camera to collect document image. Therefore, the acquisition angle is small, can not get a complete chapter image. In view of this problem, this paper designs a multi--channel camera document image stitching algorithm. In the image registration step, an area feature description method for document image is proposed to improve the feature matching efficiency. In the image fusion step, a search algorithm of the optimal splicing seam of document image is proposed to eliminate the splicing ghost. At the same time, the multi--resolution fusion algorithm is applied for optimizing the image exposure differences. According to the experimental results, this algorithm receives a good registration and fusion effect and can be better used in elec- tronic video magnify system, in which this algorithm improves the registration accuracy and speed as well as eliminates the splicing ghost and exposure differences, obtaining a wider reading angle and high image resolution.
出处
《计算机测量与控制》
2018年第2期227-231,235,共6页
Computer Measurement &Control
基金
国家自然科学基金资助项目(61401398)
关键词
电子助视系统
文档图像拼接
特征描述
最优拼接缝
多分辨率融合
electronic video magnify system
document image stitching
feature description
optimal splicing seam
multi-resolution fusion
作者简介
胥鹏鹏（1991-），男，河南周口人，硕士研究生，主要从事视频和图像处理、嵌入式系统应用方向的研究。;通讯作者：郑雅羽（1978-），男，浙江温州人，博士，副研究员，硕士生导师，主要从事视频和图像处理、深度学习及嵌入式系统方向的研究。
