一种基于视觉感知的非线性幅型比变换方法

A Non-uniform Aspect Ratio Conversion Algorithm Based on Visual Perception

随着高清电视的出现,幅型比变换的应用范围越来越广泛。本文提出了一种基于人的视觉感知特性的非线性幅型比变换方法。该方法利用运动估计技术获得视频帧的运动矢量,采用基于修正划分模糊度的模糊聚类算法对运动矢量进行聚类并估算每一帧的运动重心(MCOG)。同时针对幅型比变换后由视频序列前后帧不连续引起的视觉上的抖动现象,采用基于运动重心和中心区域的非均匀区域划分法,并针对不同的区域进行非线性比率扩展。实验结果表明,该方法很好地反映了人的视觉对视频的感知情况,变换后的视觉效果明显优于传统的幅型比变换方法。

With the emergence of HDTV, the application of aspect ratio conversion （ARC） is getting wider in related fields. In this paper, a new aspect ratio conversion algorithm is proposed based on human visual perception. To start with, the fuzzy cluster algorithm based on the modified partition fuzzy degree algorithm is applied to cluster motion vectors adaptively, and then the image motion center of gravity （MCOG） is estimated based on the clustering result. Moreover, to solve the dithering artifact, a splitting region approach based on MCOG and center region of image is developed. Finally, different regions of image are scaled with different ratios. Experimental results illustrate that the proposed method has good consistency with the human visual perception and show its superiority.