A Fast Motion Vector Estimation Method Based on Color Multidimension Projection

Abstract

An approach of fast motion estimation (ME), aiming at the stabilization problem of video surveillance system on driving vehicle, is proposed in the paper. The information of color space based on multidimension projection and brightness based on grayscale projection (GP) are jointly considered. Compared with the traditional approaches of GP, the presented approach can be applied to achieve higher precision and remain real-time performance. Furthermore, it is more robust than GP, especially when brightness distribution of video image is nearly even. Meanwhile, scaling factor estimation is also applied, which is helpful to estimate motion vector(MV) from scaling motion mode.