计算机工程

• 图形图像处理 • 上一篇    下一篇

基于多邻域信息的监控图像超分辨率算法

李云峰,李晟阳,韩茜茜   

  1. (河南科技大学 机电工程学院,河南 洛阳 471003)
  • 收稿日期:2015-05-06 出版日期:2016-06-15 发布日期:2016-06-15
  • 作者简介:李云峰(1973-),男,副教授、博士,主研方向为图像处理、计算机视觉、机器学习;李晟阳、韩茜茜,硕士研究生。
  • 基金项目:
    河南省教育厅科学技术研究基金资助重点项目“人脸图像特征提取中核张量子空间理论与方法研究”(14A413013)。

Super-resolution Algorithm of Surveillance Images Based on Multi-neighborhood Information

LI Yunfeng,LI Shengyang,HAN Xixi   

  1. (School of Mechatronics Engineering,Henan University of Science and Technology,Luoyang,Henan 471003,China)
  • Received:2015-05-06 Online:2016-06-15 Published:2016-06-15

摘要: 针对当前视频监控图像分辨率较低、目标难以辨识的问题,基于多邻域信息,提出一种监控图像超分辨率算法。根据源图像中各像素点沿多个方向灰度梯度的变化,在双三次插值的基础上,通过改进算法获得超分辨率图像的插值结果。该算法考虑了源图像中边缘部分灰度变化方向的多样性,使得重建图像的边缘细节更加自然清晰。实验结果表明,与传统双三次插值算法相比,该算法能较好地再现图像的各种边缘信息,改善图像的主观视觉效果,所获得的超分辨率图像的峰值信噪比、均方误差以及图像相似度等评价指标均优于传统算法。

关键词: 超分辨率, 双三次插值, 灰度梯度, 方向信息测度, 参数估计, 数据融合

Abstract: A monitoring image super-resolution algorithm based on multi-neighborhood information is proposed to solve the problem of target identification due to the low image resolution of current video surveillance.On the basis of the bi-cubic interpolation,according to the change of each pixel along multi-directional gray gradient in the source image,optimized algorithm is used to acquire the super-resolution image.The multiplicity of gray scale directional change in the edge part of source image is taken into full consideration,so that the details section of the reconstructed image is more natural and clear.Experimental results show that,compared with the traditional bi-cubic interpolation algorithm,the algorithm can well represent the edge information of the image,obviously improve the subjective visual effect of the imag.The evaluation indexes of the obtained super-resolution image,which include peak signal to noise ratio,mean square error,and image similarity degree,etc.,are better than traditional methods.

Key words: super-resolution, bi-cubic interpolation, gray gradient, orientation information measure, parameter estimation, data fusion

中图分类号: