Difference between revisions of "Example Video Stabilization"

From BoofCV
Jump to navigationJump to search
m
m
Line 7: Line 7:
Video stabilization attempts to overlay more recent images on top of a keyframe.  This makes it easier to identify objects which are moving relative to the background.  
Video stabilization attempts to overlay more recent images on top of a keyframe.  This makes it easier to identify objects which are moving relative to the background.  


Example File: [https://github.com/lessthanoptimal/BoofCV/blob/v0.17/examples/src/boofcv/examples/geometry/ExampleVideoStabilization.java ExampleVideoStabilization.java]
Example File: [https://github.com/lessthanoptimal/BoofCV/blob/v0.18/examples/src/boofcv/examples/geometry/ExampleVideoStabilization.java ExampleVideoStabilization.java]


Concepts:
Concepts:

Revision as of 14:15, 22 September 2014

Video stabilization attempts to overlay more recent images on top of a keyframe. This makes it easier to identify objects which are moving relative to the background.

Example File: ExampleVideoStabilization.java

Concepts:

  • Image Stitching
  • Image Stabilization
  • 2D image motion

Relevant Applets:

Related Examples:

Example Code

/**
 * Example of how to stabilizing a video sequence using StitchingFromMotion2D.  Video stabilization is almost
 * the same as creating a video mosaic and the code in this example is a tweaked version of the mosaic example.
 * The differences are that the output size is the same as the input image size and that the origin is never changed.
 *
 * @author Peter Abeles
 */
public class ExampleVideoStabilization {
	public static void main( String args[] ) {

		// Configure the feature detector
		ConfigGeneralDetector confDetector = new ConfigGeneralDetector();
		confDetector.threshold = 10;
		confDetector.maxFeatures = 300;
		confDetector.radius = 2;

		// Use a KLT tracker
		PointTracker<ImageFloat32> tracker = FactoryPointTracker.klt(new int[]{1,2,4,8},confDetector,3,
				ImageFloat32.class,ImageFloat32.class);

		// This estimates the 2D image motion
		// An Affine2D_F64 model also works quite well.
		ImageMotion2D<ImageFloat32,Homography2D_F64> motion2D =
				FactoryMotion2D.createMotion2D(200,3,2,30,0.6,0.5,false,tracker,new Homography2D_F64());

		// wrap it so it output color images while estimating motion from gray
		ImageMotion2D<MultiSpectral<ImageFloat32>,Homography2D_F64> motion2DColor =
				new MsToGrayMotion2D<ImageFloat32,Homography2D_F64>(motion2D,ImageFloat32.class);

		// This fuses the images together
		StitchingFromMotion2D<MultiSpectral<ImageFloat32>,Homography2D_F64>
				stabilize = FactoryMotion2D.createVideoStitchMS(0.5, motion2DColor, ImageFloat32.class);

		// Load an image sequence
		MediaManager media = DefaultMediaManager.INSTANCE;
		String fileName = "../data/applet/shake.mjpeg";
		SimpleImageSequence<MultiSpectral<ImageFloat32>> video =
				media.openVideo(fileName, ImageType.ms(3, ImageFloat32.class));

		MultiSpectral<ImageFloat32> frame = video.next();

		// The output image size is the same as the input image size
		stabilize.configure(frame.width, frame.height, null);
		// process the first frame
		stabilize.process(frame);

		// Create the GUI for displaying the results + input image
		ImageGridPanel gui = new ImageGridPanel(1,2);
		gui.setImage(0,0,new BufferedImage(frame.width,frame.height,BufferedImage.TYPE_INT_RGB));
		gui.setImage(0,1,new BufferedImage(frame.width,frame.height,BufferedImage.TYPE_INT_RGB));
		gui.autoSetPreferredSize();

		ShowImages.showWindow(gui,"Example Stabilization");

		// process the video sequence one frame at a time
		while( video.hasNext() ) {
			if( !stabilize.process(video.next()) )
				throw new RuntimeException("Don't forget to handle failures!");

			// display the stabilized image
			ConvertBufferedImage.convertTo(frame,gui.getImage(0, 0),true);
			ConvertBufferedImage.convertTo(stabilize.getStitchedImage(), gui.getImage(0, 1),true);

			gui.repaint();

			// throttle the speed just in case it's on a fast computer
			BoofMiscOps.pause(50);
		}
	}
}