We run our denoising algorithm on the videos from [13] using both our flow and their flow. These videos were not captured using motion-based exposure control.
Top: Our denoising result using our flow. Bottom: Our denoising using the flow method from [13]. In this example, our method benefits from [13]'s flow to achieve better temporal consistency (bottom) that is comparable to [13], since the motion is relatively small in this video and the flow method used in [13] provides more temporal consistency in smooth regions than our flow method.
Top: Our denoising result using our flow. Bottom: Our denoising result using the flow method from [13]. In this example, our method benefits from [13]'s flow to achieve better temporal consistency (bottom) that is comparable to [13], since the motion is relatively small in this video and the flow method used in [13] provides more temporal consistency in smooth regions than our flow method.