Improved super-resolution reconstruction algorithm of non-local mean video
-
摘要: 针对基于非局部均值(NLM)的视频超分辨率重建方法存在结果过于平滑、收敛速度慢及计算量大等问题,提出一种改进的NLM视频超分辨率重建算法。该方法采用模糊边缘补足算法将经过预处理的视频图像分成平坦区和纹理区;对于平坦区,采用直方图均衡化的方式进行图像增强处理,以减少算法计算量;对于纹理区,采用改进的NLM重建算法进行处理,通过设计多方向自适应搜索窗并引入邻域相干系数修正相似性权值,以增强重建图像的纹理细节,加快算法的收敛速度;将重建的纹理区与增强的平坦区进行叠加归一化处理,完成整个视频图像的超分辨率重建。实验结果表明,该算法能够在提高重建图像纹理细节及峰值信噪比的同时,降低算法的整体复杂度,缩短重建时间。Abstract: In view of problems of video super-resolution reconstruction method based on non-local mean (NLM) that reconstructed image was too smooth, convergence speed was slow and calculation amount was large, an improved super-resolution reconstruction algorithm of non-local mean video was proposed. The method uses fuzzy edge complement algorithm to divide preprocessed video image into flat region and texture region; for flat region, image enhancement processing is performed by histogram equalization to reduce the amount of algorithm calculation; for texture region, it is processed by the improved NLM reconstruction algorithm, and similarity weights is corrected by designing a multi-directional adaptive search window and introducing neighborhood coherence coefficients, so as to enhance texture details of the reconstructed image and speed up the convergence of the algorithm; Superimposed normalization of the reconstructed texture region and the enhanced flat region is performed to complete super-resolution reconstruction of the entire video image. The experimental results show that the proposed algorithm can reduce overall complexity of the algorithm and shorten reconstruction time while improving the texture details and peak signal-to-noise ratio of the reconstructed image.
-
-
期刊类型引用(4)
1. 代晓亮,孙建维,赵帅,刘发义,张豪. 胡底煤矿强矿压微震监测预警技术研究. 建井技术. 2024(05): 11-16 . 百度学术
2. 张向阳,蔡保华,焦伟,马江鹏,尚楠,余涛涛,马文涛. 深部综放工作面冲击危险性分析及防冲技术研究. 煤炭科技. 2023(05): 99-104+109 . 百度学术
3. 刘华博,段翠芳,孟凡净,花少震. 单侧采空工作面矿震活动规律与冲击地压相关性分析. 河南工学院学报. 2022(01): 1-5 . 百度学术
4. 陈玉涛. 邻正断层不规则孤岛工作面冲击危险性评估方法研究. 矿业安全与环保. 2022(06): 118-122+129 . 百度学术
其他类型引用(1)
计量
- 文章访问数: 110
- HTML全文浏览量: 11
- PDF下载量: 13
- 被引次数: 5