摘要
针对视频图像普屏到宽屏的转换过程中主景图像出现扭曲变形的问题,提出了一种非线性区域位置根据兴趣区域检测结果变化的非线性缩放算法.该算法解决了现有方法只能保证图像的中间区域不失真而靠近图像两边的主景区由于插值点密度的急剧变化发生大的扭曲和变形的问题.算法中的兴趣区域的检测以人物为主,采用了肤色检测和Harris角点检测相结合的方法.该算法已通过电路设计实现,同时经过兴趣区域特征圆度分析表明,此种处理方法的结果可以使主景区扭曲较小.
To solve the problem of main spot distortion when normal video signal is displayed on wide screen devices,a novel method of nonlinear scaling was proposed,in which the locations of nonlinear parts change with interest-region detection results.This method overcame the disadvantage of existing nonlinear scaling methods that only cen-tral region is not distorted but main spot near the border may be distorted because of the great change in interpolation point density.The interest-region detection mainly searched for figure by skin tone and Harris corner detection.The method was realized in hardware,and through analyzing roundness of interest-region characteristic circle,it was proved that the main spots after being processed by the proposed method were less distorted than those processed by other methods.
出处
《天津大学学报（自然科学与工程技术版）》
EI
CAS
CSCD
北大核心
2012年第7期635-640,共6页
Journal of Tianjin University：Science and Technology
基金
国家自然科学基金资助项目(61076024)
教育部博士点新教师基金资助项目(200800561112)
关键词
非线性插值
肤色检测
HARRIS角点
特征圆度
nonlinear interpolation
skin tone detection
Harris corner
roundness of characteristic circle
