一种结合相位相关法和AKAZE算法的纹理缺乏地区影像匹配技术

An Image Matching Technique Combined with Phase Correlation Method and AKAZE Algorithm for Texture Lack Region

由于纹理缺乏地区影像缺少鲜明特征,特征重复性严重,使得单纯的基于特征的匹配算法很难获得唯一性较强的特征描述符,导致同名点误匹配的比率大大增加。AKAZE算法以非线性尺度空间为基础,较好地顾及了影像的纹理结构及细节信息,对纹理缺乏地区的影像匹配具有一定的应用价值。基于频域的Fourier-Mellin相位相关技术,通过对影像的频域及对极数变换,计算互功率谱可获得影像间的旋转、尺度和平移参数。因此,本文通过Fourier-Mellin变换计算影像间的几何关系,对AKAZE算法检测出的特征点进行了匹配范围的约束,初步解决了纹理缺乏地区的影像匹配问题。试验结果表明,该方法可以提高匹配的正确率及同名点对数。

Because of the lack of texture region is short of distinctive features or having serious repetitive features, the matching algorithm purely based on feature is very difficult to obtain the strong unique feature descriptors, which leads to greatly increasing of mismatching ratio. But the AKAZE algorithm based on the nonlinear scale space, which can better take into account the texture and detail information of image, so it has a certain application value for image matching in the lack of texture region. Calculating the cross- power spectrum between images by Fourier-Mellin phase correlation technique using the FFT and log-par conversion, can obtain the rotation, scale and translation parameters between the two images. So, this paper obtains the geometric relationship between the two images by Fourier-Mellin conversion, which is used for restraining the matching range of the feature points detected by the AKAZE algorithm, to solve the problem of image matching in lack of texture region. Experimental results show that this method can increase the number of matching points, and the matching accuracy.