基于压缩采样的图像序列光流场计算

Optical Flow Calculation of Sequential Images Based on Compressive Sampling

压缩采样是一个新兴的研究领域,为亚奈奎斯特采样频率进行图像获取提供了一个框架。光流场计算在计算机视觉系统中应用广泛,将压缩采样理论引入图像序列的光流计算问题,提出利用梯度图像的空域稀疏性和傅里叶变换的时移性,直接由测量数据求解光流场,该方法无需重建原图像序列,且在同样检测性能的情况下,所需测量数据少,可以降低采样、通信和存储成本。

The compressive sampling provides a new frame for image acquiring under sub-Nyquist sampling rate.The optical flow calculation is widely used in computer vision system.In this paper,the compressive sampling theory is introduced to solve the optical flow calculation.Instead of using reconstructed original images,the optical flow field can be obtained by exploiting the spatial sparsity of gradient images and time shift feature of Fourier transform.The experiment results on real image sequence show that,under the same condition,this method is more efficient and needs less costs for sampling,communication and storage.