Blackmagic Cintel Scanner User Manual  Object Removal

The object removal plugin is best used in the ‘color’ page, and uses the DaVinci Neural Engine to attempt to remove an object in the frame as automatically as possible. This plugin works best when removing a moving object that passes over a temporally stable background, or dirt on the lens of a shot where the camera is in motion. Smaller objects get better results than larger objects, but your results really depend on the footage. Here’s a simple procedure that shows how to do this.

To remove a moving object from a clip:

1. In this example, a drone is flying through a long shot that’s being simultaneously recorded. We’ll remove the drone using a window to identify the feature to be removed using the object removal plugin. The original shot with a drone that needs to be removed In simple cases, it’s often easiest to apply the object removal effect to a corrector node, so you can use a window or qualifier within that node to isolate the feature you want to remove. That’s what we’ll do in this example.

2. Use the Window palette to draw a window around the object that needs to be removed. You’ll get the best results using windows or masks that hug the feature being removed fairly closely.

3. Track or keyframe the window to move with the feature you’re removing. Again, you’ll get better results the closer your window hugs the object being removed, and it’s good to have some softness at the edge of this window. The object that needs to be removed is isolated with a window

4. Drag and drop the object removal plugin onto the node in which you’ve just isolated the feature to be removed.

5. The ‘use OFX alpha’ option will be activated automatically in the node’s contextual menu to enable the object removal plugin to use whatever key has been created within that node to do its work.

6. Click the ‘scene analysis’ button, and wait for the analysis to finish. If the object you’re removing is moving but the camera is locked, you can turn on the ‘assume no motion’ checkbox to improve your results in this case.

If your footage is ideal for object removal, the object will disappear once analysis is complete, replaced by a seamless background derived from detail found on neighboring frames.

The result after object removal analysis is complete

The object removal plugin is highly footage dependent, and you won’t always get this good a result this easily. Problems with the result are shown via gray, either gray fringing or solid gray filling the replacement window. Gray shows you where the current settings are failing to find background content with which to fill in the patch you’re removing. If this happens, there are two things you can try.

• If you notice while playing through the analyzed result that the object removal mask has gray fringing on some frames, you can try adjusting the ‘search range’ slider, which is the distance, in frames, from the current frame that the object removal plugin is searching for replacement image detail. For example, if the search range is 20, it searches +/-20 frames from the current location, or 40 frames total. The allowance of 10 frames means we look at every 4th frame. You will generally get the best results for the smallest range that gives an acceptable result.

• If you’re noticing that the object removal mask is filled entirely with gray on some frames, this means that background fill couldn’t be easily generated for those frames. In this case, you can try clicking the ‘build clean plate’ button, which takes a ‘best guess’ approach to generating a background with which to fill the frame and integrates this with frames that could be successfully filled in.

• If the patch is successfully filled, but the result isn’t blending well with the background, you can try changing the blend mode. The default is ‘linear’, which is a simple cloning operation, but you can also choose ‘adaptive blend’, which can provide better results except in certain situations where the edges of the replacement patch have a different color or brightness than the background.

• The ‘scene mode’ menu provides different methods of analyzing the scene, for improving the analysis of how the area that needs to be replaced moves, to best determine how to fill the hole left by the object being removed. Background analyzes the entire image except for the object region. Boundary analyzes the boundary area surrounding the object region. Object is for analyzing an object that moves with the background, like a sticker that’s on a window while the camera moves.