A new motion estimation method for motion-compensated frame interpolation using a convolutional neural network
Giyong Choi, PyeongGang Heo, Se Ri Oh, HyunWook Park · 2017
The motion-compensated frame interpolation (MCFI) methods usually use block matching algorithms (BMAs) for motion estimation (ME). However, the conventional BMAs that are originally developed by minimizing the prediction errors often fail to project the object motion. In this paper, we present a new MCFI method that utilizes a convolutional neural network (CNN) to find the motion vector (MV) with reliability. The CNN model which is used to estimate MVs is trained to track the projected object motion as closely as possible. Experimental results using the standard test video sequences show that our proposed ME method acquired more reliable MVs than conventional ME methods. Furthermore, our proposed MCFI method improves the average peak signal-to-noise ratio (PSNR) of interpolated frames.