基于时空特性自适应插值的混合去隔行方法
Hybrid De-Interlacing Algorithm with Adaptive Interpolation Based on Temporal-Spatial Characteristics

为了实现隔行扫描到逐行扫描的视频格式转换, 需要使用去隔行插值技术．针对传统运动补偿插值算法的误差传递、运动估计不准、计算复杂度较高等缺陷, 提出一种基于奇偶一致性搜索的运动补偿去隔行方法; 针对传统基于边缘方向插值算法的边缘定位不准等缺陷, 提出一种基于梯度辅助的场内多方向插值方法．将上述两种方法与场合并插值和行平均插值相结合, 提出一种新的基于时空特性自适应插值的混合去隔行方法．该方法能够利用视频中的时空特性, 有效地解决传统去隔行算法因运动估计错误和边缘定位不准导致的插值失真等问题．实验结果表明, 与传统去隔行算法相比, 该方法能够在降低总体计算复杂度的情况下获得0.7 dB左右的PSNR提升．
De-interlacing is a key technology in the process of converting a interlaced video into a progressive video. But the traditional motion compensated de-interlacing algorithms often suffer from some troubles，such as error propagation，inaccurate motion estimation and high computation complexity. Thus，a motion compensated de-interlacing algorithm based on parity consistency search was proposed. In addition，in view of the problem that the edge direction wasn’t detected exactly in the traditional de-interlacing algorithms based on the edge direction，a multi-direction interpolation algorithm based on gradient-assistance was proposed in this paper. Finally，the above proposed de-interlacing algorithms，field merging interpolation and line averaging interpolation were merged adaptively to form a hybrid de-interlace with adaptive interpolation based on temporal-spatial characteristics. Making use of the temporal-spatial characteristics of the video，the finally proposed algorithm effectively solves the problems that the wrong motion estimation and the inaccurate edge direction resulted in mistaken interpolation. The experimental results show that compared with traditional de-interlacing algorithms，the proposed algorithm could offer 0.7 dB PSNR gain under low computational complexity