一种基于功率谱估计的图像相位匹配方法

摘要

相位相关法作为常用的相位匹配技术在使用中会遇到相位卷绕问题,一般频域求解方法需要进行相位解绕和最小二乘或RANSAC拟合,但相位解绕过程繁琐,且最小二乘拟合法易受噪声干扰,RANSAC拟合方法时间开销大.为此,提出了一种基于平均周期图法功率谱估计的相位匹配方法,将空间域问题转化到频域内求解,并成功规避了相位解绕步骤,相位匹配精度达到了亚像素量级.与经典频域分析方法对比,方法可在规避相位解绕的同时获得精度相当的有效结果.最后对方法的分辨精度进行了仿真分析,实验结果证明能够达到亚像素级的视差分辨率,为相位匹配在频域中的研究提供了新的思路.%As the commonly used phase matching technology,the phase correlation method will encounter phase wrapping problems.Thus,the general method in the frequency domain contains phase unwrapping and least squares fitting or RANSAC fitting.However,the phase unwrapping process is tedious,and the least square fitting method is vulnerable to noise interference,as well as the RANSAC fitting method is time consuming.An analysis of the phase matching method is proposed based on the power spectrum estimation using averaged periodogram method,transforming into the frequency domain to solve spatial domain problem and successfully avoiding the phase unwrapping steps.And the phase matching accuracy reaches the sub-pixel level.By comparing with the classical analysis method in frequency domain,this method can obtain the equivalently accurate results of the phase unwrapping method.Finally,the resolution accuracy of the method is simulated and analyzed.The results of experiment show that this method can reach the sub-pixel disparity resolution,which provides a new idea for the research of phase matching in frequency domain.