局部相位相关法在密集匹配中的应用及精度分析

Application and Precision Analysis of Local Phase Correlation in Dense Matching

传统相位相关算法广泛应用于图像配准与拼接中。本文通过对双目立体像对的加窗二维傅里叶变换,采用局部相位相关法获取互相关功率谱,实现图像局部相位相关,得到图像窗口间的亚像素级平移参数,进而获得窗口中心处的视差值。试验结果表明,在选择合适的窗口尺寸下,该算法匹配精度较高,且对随机噪声和光照变化具有较强的稳健性。

Traditional phase correlation algorithm is widely used in image registration and mosaicking. Based on the windowed Fourier transform of binocular stereopair,this paper realized the local phase correlation in image,and got the sub-pixel translation parameters between the windows by obtaining the cross-correlation power spectrum using the local phase correlation method,so as to calculate the disparity at the center of the window.The experimental results show that the proposed algorithm with appropriate window size has higher matching accuracy and strong robustness to random noise and illumination changes.