用小波锥分解提取旋转和尺度不变特征

USING WAVELET TRANSFORM TO EXTRACT ROTATION AND SCALE INVARIANTS

利用极坐标下Fourier变换的模所具有的旋转不变性以及Mellin变换的模所具有的尺度不变性,经过小波锥分解,提取其高层低频分量作为特征进行模式识别。对于0～9的十个不同旋转角度和尺度大小的160幅图像都能正确的进行识别。

In this paper the Fourier transform and Mellin transform are used to obtain the rotation and scale invariant characteristic parameters. In order to reduce the dimension of the characteristic space, wavelet transform is used on the above characteristic space to ex-tract the compact characteristic templates. The algorithm works well for the test image set with 10 digits including different orientations and scales.