摘要
提出了一种从隔行扫描到逐行扫描的格式转换算法。该算法首先计算相隔一场的两个电视场亮度信号之间的差值,即相邻的两个奇场(或偶场)亮度信号之间的隔场差。然后,对差值信号执行形态开运算,以滤除一些高频噪声。通过使用隔场差,不仅能正确检测静止区域和运动区域,而且还能检测经3 2下拉变换后的电影序列。对于检测到的电影序列,则进行3 2下拉恢复处理,输出电影模式的逐行扫描帧;否则,采用运动自适应的插补方法进行去隔行处理,转换成具有高垂直分辨率的逐行扫描帧图像。
This paper proposes a de-interlacing algorithm. First, it calculates the luminance difference between the forward field and the backward field, i.e., the luminance difference between two successive odd fields (or two successive even fields). Then, the morphological opening is operated to eliminate the high frequency noise in the difference field. The forward-backward difference can not only detect the static and motion regions but also detect the 3-2 Pull-Downed movie sequences. If it is movie sequences, we should adopt the 3-2 Pull-Downed recovery to perfectly reconstruct the progressive frame; Otherwise, we use the motion-adaptive interpolation to get high vertical resolution progressive frames from interlaced frames.
出处
《南京邮电学院学报(自然科学版)》
2004年第3期53-57,共5页
Journal of Nanjing University of Posts and Telecommunications