论文部分内容阅读
光流是指图像亮度模式的表现运动，通过光流我们可以了解物体的许多重要的运动特征，在运动图像计算和分析中光流扮演着十分重要的角色。本文研究了全局平滑和局部平滑对光流计算的影响。为提高光流估计的精度和效果，本文提出了一种基于全局和局部平滑的光流计算的方法，首先对图像进行结构纹理分解，考虑全局和局部平滑项，得到欧拉-拉格朗日方程，在多尺度的框架下进行迭代求解，对迭代过程中计算的光流使用带权值的中值滤波进行平滑，最后用 guided 滤波平滑输出的光流。实验结果表明本文方法能取得很好的结果。“,”Optical flow is the distribution of apparent velocities of movement of brightness patterns in an image. Through the optical flow, we can understand many important moving characteristic of the object, it plays a very important role in the calculation and analysis of motions.We study the effect of local smoothness and global smoothness for optical flow. In order to improve the accuracy and effectiveness of flow estimation, we propose an optical flow method based on global and local smooth:first we process structure-texture decomposition for input images, then obtain Euler - Lagrange equation considering the global and local smoothing term, using an iterative scheme to solve the optical flow under the multi-scale framework, smooth flow in the iterative process of calculation with the weighted median filter, and finally smooth the output of the optical flow using the guided filter.The results demostrate that our proposed method can achieve impressing results.