Fourier-Mellin矩在图像识别中的应用

摘要

为了验证Fourier—Me｜lin矩图像识别中的识别能力，本文研究了其在两种坐标下的计算和重建效果、抗噪性试验。在笛卡尔坐标下，图像重建直接计算，不必转换为极坐标，避免极坐标在了转换时产生的几何和计算误差，试验表明：笛卡尔坐标下，Fourier—Mellin矩的重建比极坐标下更精确，并且OFFM矩对噪声有很好的鲁棒性。%In order to test the ability of Fourier-Mel.lin moment in image recognize, research work are done that two ways of computation and image reconstructed in a polar coordinate and in a Cartesian coordinate system, noise robustness are implemented in this paper. The Fourier-Mellin moment can be computed directly in a Cartesian coordinate, and this method completely removes both approximation and geometrical errors produced by the conventional methods. The experiments show that the reconstructed image by the Fourier-Mellin moment in a Cartesian coordinate is more accurate than that in a polar coordinate and the Fourier-Mellin moment has strong robustness for noise images.