Motion Field and Occlusion Time Estimation via Alternate Exposure Flow

This paper presents an extension to optical flow-based motion estimation using alternating short- and long- exposed images. While traditional optical flow algorithms rely on consecutive short-exposed images only, long-exposed images capture motion directly in the form of motion blur. This additional information can be used to achieve more robust and accurate motion field estimation as well as to extract the moment of occlusion. We introduce an image formation model that relates the long-exposed image to its preceding and succeeding short-exposed images in terms of dense 2D motion and per-pixel occlusion/disocclusion timings. Based on this image formation model, we describe a practical algorithm to estimate the motion field not only for completely visible image regions but also for pixels becoming occluded. For these pixels the Alternate Exposure Flow (AEF) also determines the moment of occlusion. We describe the application of AEF in frame interpolation to demonstrate the advantage of the additional long exposure information.