运用独立元分析的仿射运动估计算法
Estimation of Affine Motion by Using Independent Component Analysis
-
摘要: 目前在视频流中的运动估计都是针对平移运动的,对于旋转和扭曲运动效果很差,采用Oirrak等提出的Fourier方法可实现旋转和扭曲运动估计,但当目标轮廓简单或噪声大时效果不佳. 提出用独立元分析(ICA)方法对图像目标的扭曲运动进行估计,该方法对噪声具有很好鲁棒性,克服了Fourier方法对简单物体仿射参数估计效果差的局限性;同时,采用快速ICA算法计算量小,在视频处理和通信领域中有很好的应用前景.Abstract: Traditional motion estimation methods in video were based on translation,and it was ineffective for rotation and twisting.Methods based on Fourier descriptor proposed by Oirrak et al could estimate the affine transformation,but its performance degraded for simple contour or contour with great noise.New approach with independent component analysis (ICA) is robust to noise,and free from the shortcomings of Fourier method.In addition,the proposed approach uses a fast ICA algorithm with higher computation speed.
下载: