A Novel Framework for Frame Rate Up Conversion by Predictive Variable Block-Size Motion Estimated Optical Flow

Abstract

This paper proposes a novel framework for frame rate up conversion. The framework contains a motion field estimator employing a hybrid of a novel predictive variable blocksize motion estimation and robust optical flow algorithm. This motion field is more robust and accurate than pure optical flow or pure motion estimation methods. Therefore, in-between frames can be estimated more accurately by the new hybrid method. After the motion field estimation, pixel interpolation following motion field interpolation is done for in-between frames. It is shown the proposed method has better than pure motion estimation and pure optical flow methods in terms of both subjective and objective measurements.