Frame Rate Up-Conversion for HDR Video Using Dual Exposure Camera

In this paper, we propose a new frame rate up-conversion (FRUC) for HDR video using a dual exposure camera that captures alternating frames with two different exposure levels (short and long). First, optical flows are estimated in the proposed convolutional neural network (CNN)-based motion estimation (ME) analyzing dual-exposed frames. Then, with the estimated true motions and motion occlusion maps, an HDR frame is interpolated while suppressing ghost artifacts caused by HDR fusion. Finally, FRUC is performed by re-using motions and occlusion maps. Experimental results indicate that the proposed method is effective to eliminate ghost artifacts and increase frame rate at the same time.