A data fusion solution to the accuracy-efficiency trade-off problem in motion estimation

There is often a trade-off between the accuracy and the speed of optical flow techniques. Given similar computational resources, this trade-off results in some techniques making infrequent but accurate estimates of motion while others make more frequent but less accurate estimates. This paper proposes a novel method which enables the results of the more accurate techniques to be used to improve the accuracy of the faster techniques. The method is a combination of data fusion methods with predictors