Digital image stabilization based on variance properties

Abstract

In this paper, a novel method for the digital image stabilization is proposed that use variance properties(VDIS) to reduce the computational cost of block matching algorithm for local motion estimation. This method based on calculation of variance of pixels in each frame and accumulated variance of each pixel in each location from first reference frame to current frame. Then calculate summation of all accumulated variance for current frame and using them for select best macroblock. Four macroblocks are used in each frame. By statistical calculation the best macroblock is selected then local motion vector(LMV) of selected macroblock, is estimated by using Full search algorithm. LMV of selected macroblcok is used as global motion vector(GMV) and it is used to stabilize that frame. Full search algorithm is used for motion estimation in selected macroblock, but partial distortion elimination (PDE)[1] method was used to terminate the improper candidate blocks and reduce computation for block matching.