An improved optical flow estimation method based on CLG-TV model

The combining method of neighborhood and non-neighborhood is proposed to be used for amending the computed results of optical flow, which has improved the robustness of flow estimation. This method uses the estimated frame based on CLG-TV model, carries out the decomposition of structure and texture for the images, proceeds weighted approach for data term and smooth term, uses coarse-to-fine method to conduct iterative solution, corrects the flow calculated during iterative process and uses smooth output flow of non-local median filtering. The experimental result shows that, the improved algorithm that has been proposed possesses better robustness and its accuracy for optical flow estimation is better than the accuracy of CLG-TV algorithm.