Robust point matching is a critical and challenging process in feature-based image registration. In this paper, an invariant feature point matching algorithm is presented by introducing the Polar Complex Exponential Transform (PCET), a new kind of orthogonal moment reported recently. Similar to orthogonal complex mo-ments, PCETs is defined on a circular domain. The magnitudes of PCETs are invariant to image rotation and scale. Furthermore, the PCETs are free of numerical instability, so they are more suitable for building shape descriptor. In this paper, the invariant properties of PCETs are investigated, and the accurate moments are selected elaborately. During similarity measurement， the cross correlation function is reconstructed by in-variant PCET moments (IPCETs) combining both the magnitude and phase coefficients and maximized to match the control- point pairs. Then the most “useful” matching points that belong to the background are used to find the global transformation parameters between the frames using the projective invariant. The discrimina-tive power of the new IPCETs descriptor is compared with major existing region descriptors (complex mo-ments, SIFT and GLOH). The experimental results, involving more than 10 million region pairs, indicate the proposed IPCETs descriptor has, generally speaking, produce a more robust registration under photometric and geometric performances.