Fast Global Motion Estimation in Two Sampling Steps

An important technique in image and video processing is global motion estimation (gme). the common gmemethods can be classified in direct and indirect categories. whereas the direct global motion estimation techniquesboast reasonable precision they tend to suffer from high complexity. as with indirect methods, though presentinglower complexity, they mostly exhibit lower accuracy than their direct counterparts. in this paper, the authorsintroduce a robust algorithm for gme with near identical accuracy and almost 50-times faster than mpeg-4verification model (vm). this approach entails two stages in which, first, motion vector of sampled block is employedto obtain initial gme then levenberg-marquardt algorithm is applied to the subsampled pixels to optimize the initialgme values. as will be shown, the proposed solution exhibits remarkable accuracy and speed features withexperimental results distinctively bearing them out.