Computer Engineering and Applications ›› 2012, Vol. 48 ›› Issue (9): 178-180.

• 图形、图像、模式识别 • Previous Articles     Next Articles

Robust and real-time approach for super-resolution image reconstruction

LIU Rundan1, PAN Xinsheng2   

  1. 1.Yangjiang Vocational and Technical College, Yangjiang, Guangdong 529566, China
    2.Guangzhou Xingbo Information Technology Co., Ltd, Guangzhou 510630, China
  • Received:1900-01-01 Revised:1900-01-01 Online:2012-03-21 Published:2012-04-11

一种实时鲁棒的超分辨率图像重建方法

刘润丹1,潘新生2   

  1. 1.阳江职业技术学院 机电系,广东 阳江 529566
    2.广州星博信息技术有限公司事业三部,广州 510630

Abstract: In intelligent transportation systems, the super-resolution reconstruction is widely used. In order to improve the quality of super-resolution image, a new approach for super-resolution image reconstruction is proposed. This algorithm extracts all unique and robust feature descriptors of SURF from low resolution images, along with the matching calculation. The least square method is used to estimate motion parameters. Image interpolation is implemented by ANC algorithm. Experimental results show that this approach is characterized by robustness and real time, and the quality of reconstruction result is better than that of other reconstruction algorithms.

Key words: super-resolution, image reconstruction, Speeded Up Robust Features(SURF), adaptive normalized convolution

摘要: 在智能交通系统中,超分辨率重建技术有着广泛的应用。提出一种新的超分辨率图像重建方法,用于提高超分辨率图像的重建质量。该方法从低分辨率图像中提取出具有独特性和鲁棒性的SURF特征描述子,进行匹配计算,采用最小二乘法估计运动参数,采用ANC算法对图像进行插值计算。实验结果表明,该方法具有较强的鲁棒性和实时性,重建质量较之其他重建算法要高。

关键词: 超分辨率, 图像重建, 快速鲁棒性特征(SURF), 自适应归一化卷积