\r张 涛¹，潘希豪¹，王 昊¹，雷志春²

AuthorsHTML:\r张 涛¹，潘希豪¹，王 昊¹，雷志春²

AuthorsListE:\rZhang Tao¹，Pan Xihao ¹，Wang Hao¹，Lei Zhichun ²

AuthorsHTMLE:\rZhang Tao¹，Pan Xihao ¹，Wang Hao¹，Lei Zhichun ²

Unit:\r1. 天津大学电气自动化与信息工程学院，天津 300072；
2. 天津大学微电子学院，天津 300072


Unit_EngLish:\r1. School of Electrical and Information Engineering，Tianjin University，Tianjin 300072，China；
2. School of Microelectronics，Tianjin University，Tianjin 300072，China

Abstract_Chinese:\r目前，隔行扫描广泛应用于电视广播系统中，去隔行处理普遍采用多算法融合的方法，由于空域插值和时域插值的不准确性以及各算法之间结合权值的不合理性，导致融合方法的去隔行处理效果并不理想．因此，本文在分析中心独立插值算法和垂直时域滤波算法优缺点的基础上，提出了一种基于时空域方向自适应结合的去隔行方法．本方法综合考虑场内相邻像素之间的相似性和像素位置之间距离的关系，结合不同场之间的时域信息，使用更为合理的权值自适应地将空域插值和时域插值结合在一起，以增强去隔行方法结果的准确性和稳定性．本方法基于双边滤波器的原理，以不依赖中心像素的方式进行空域插值，采用垂直时域滤波算法，选择合理的滤波系数进行时域插值．仿真实验结果表明，本文所提的方法和现有去隔行方法相比较，在主观和客观上均有明显的改善．主观上与目前主流的去隔行方法相比，该方法能够在一定程度上保护边缘，减少边缘模糊，处理后的图像质量较好．客观上与其他方法相比，本文所提方法的PSNR 值均高于其他算法，计算复杂度上，本文所提方法在运算操作数和运算时间上也相对较低．综合算法的图像质量评价和计算复杂度而言，本文所提方法的图像质量较高，计算复杂度相对较低，比较适合于实时系统．

Abstract_English:\rPresently，interlaced scanning has been employed in most current broadcasting systems．Although fusion methods are commonly used for de-interlacing，they are not ideal，resulting in poor performance due to the inaccuracy of spatial interpolation，the inaccuracy of time-domain interpolation，and the unreasonableness of combining weights among algorithms．In this paper，we propose a de-interlacing method based on center-independent interpolation and vertical temporal filtering adaptively combining time-domain and space-domain directions．This method takes the similarity and the distance between adjacent pixels in the field into account，combines the time-domain information between different fields，and uses more reasonable weights to adaptively combine the spatial and time-domain interpolations．The result of the proposed method was more accurate and stable．This method is based on the principle of bilateral filtering．It does not rely on the central pixel，adopts the vertical time-domain filtering algorithm，and selects more reasonable filtering coefficients．Experimental results showed that the proposed method provides noticeable improvements over existing methods in terms of both subjective and objective evaluations．Subjectively，the proposed method had the best visual quality with fewer artifacts in the textural details．Objectively，the proposed method had larger PSNR than those of other algorithms，and low computational complexity with respect to operation number and operation time．Therefore，the proposed method is more suitable for real-time systems due to its better image quality and lower complexity．

Keyword_Chinese:去隔行；双边滤波；中心独立插值；垂直时域滤波；自适应结合

Keywords_English:de-interlacing；bilateral filtering；center-independent interpolation；vertical temporal filter；adaptively combination


PDF全文下载地址:http://xbzrb.tju.edu.cn/#/digest?ArticleID=6345