摘要
为实现隔行扫描到逐行扫描的视频格式转换,文中提出了一种时空权重和边缘自适应的去隔行算法,主要包括4个同极性场运动估计、小角度边缘搜索、时空权重计算、自适应插值等.运动估计通过比较同极性场的像素块绝对值差和(SAD)与运动阈值实现运动估计,并且利用前一场的运动信息进行鲁棒性处理,提高运动检测的精度.小角度边缘搜索采用自适应搜索半径的方法实现最小6°的小角度边缘检测.最后,通过时空权重自适应的插值算法实现去隔行处理,取得很好的处理效果.
The function of de-interlace is to convert interlaced images to progressive ones. In this paper, a de-interlacing algorithm based on spatio-temporal weight and edge direction is presented. It consists of 4 same-parity fields motion estimation, low-angle edge detection, spatio-temporal weight calculation, and adaptive interpolation. Same-parity fields are used to improve the accuracy of motion estimation. It estimates the motion by comparing the sum of absolute block difference （SAD） between 4 fields with the threshold of motion. And it detects low-angle edge using an adaptive searching radius in which the 6°edge can be detected. Experimental results show that the proposed method achieves high image quality and low hardware complexity.
出处
《计算机学报》
EI
CSCD
北大核心
2007年第4期655-660,共6页
Chinese Journal of Computers
基金
江苏省专用集成电路设计重点实验室开放研究课题(JSICK0405)资助.
关键词
运动估计
边缘搜索
时空权重自适应插值
去隔行
motion estimation
edge detection
spatio-temporal adaptive interpolation
de-lnterlace
