An image matching algorithm based on contourlet transform

Wavelet transform can't utilize the geometry feature of data fully, and was not the most optimal representation of high dimension function. Aiming at the shortcoming of wavelet transform, a new image matching method was proposed based on the Contourlet transform, which was one of the multiscale geometric analysis methods. The original image was decomposed by the Contourlet decomposition firstly and the low-frequency subimage and band-pass subimage are obtained. The Hausdorff distance and projection entropy was used for low-frequency subimage to get the rough matching results, the position of rough matching results was corresponded to the original image and then the fine matching results were obtained by the normalized product correlation similarity measure. The high precision and quickness of the proposed method is verified by the simulation results.