High Dynamic Range Video through Fusion of Exposured-Controlled Frames

In this paper, we present a method for generating high dynamic range (HDR) video through fusion of exposure-controlled frames in a stationary video camera system. Frames captured by commercial video cameras have a limit dynamic range, even if we use auto-exposure control (AEC), and these frames often show saturation areas which may degrade the image quality. To overcome these, in the video sequence alternating short and long exposures, we combine neighboring frames using an image fusion approach, since images under different exposures have different visibility in a scene. When we fuse exposure frames, there are two considerations; controlling exposures and handling moving objects. For the first, we control short and long exposures with the AEC approach. For the other, we adjust fusion masks to reduce ghosting artifacts by moving objects. Our method has the advantage of automatically selecting proper exposure times to make outputs more visible than fixed exposure ones. It also reduces motion blur distortion caused by moving objects in output frames, by using the maskadjustment method which is simpler and faster than frame registration methods.To verify the proposed method, we show experimental results using a real scene.