基于棱边方向检测与运动补偿的视频去隔行算法

New De-interlacing Algorithm Based on Edge-Dependent Interpolation and Motion Compensation

提出了一种基于棱边方向检测与运动补偿相结合的去隔行算法。首先通过Canny算子提取出棱边,将图像区分为棱边和非棱边两部分。对于棱边丢失的部分,分别在空域内沿棱边方向和时域内沿棱边移动方向插值,然后利用中值滤波得到丢失的棱边信息;对于非棱边部分,采用多级相位相关运动估计法,对当前场与前后两场计算出相应的运动矢量,然后在时域和空域中利用中值滤波计算出丢失的信号。实验结果表明,本文方法优于一般的去隔行算法,如线性空域插值等。同时本文提出了两个去隔行评价标准,可以定量检验去隔行方法的效果。

A new method for converting interlaced video into a progressively scanned video and image is introduced. The new method combines an edge dependent interpolation （EDI） algorithm with the motion compensation. The Canny operator is used to extract edges of the image and to partition an image into two parts： edge and non-edge areas. The missing pixels in the edge area are interpolated along the edge direction and the edge shift direction. The missing pixels in the non-edge area are interpolated along the motion direction combined with the spatio-temporal data. The hierarchical phase-correlation method（PCM） is used to find the motion vectors between the present field and the past one, and between the present field and the future one. The improved results are superior over the conventional de-interlacing methods in linear spatial interpolation. Two criteria are used to evaluate the performance of the de-interlacing algorithm.