Example Detect Interest Points

From BoofCV
Revision as of 09:46, 9 November 2015 by Peter (talk | contribs)
Jump to navigationJump to search

Interest points are a general term in computer vision for points in the image that can detected and are relevant for higher level processing. Interest points are commonly used by image stabilization and structure from motion applications to track how the image changes from frame to frame. The following example shows how interest points can be detected easily using the InterestPointDetector<T> interface.

InterestPointDetector is a generalized interface that allows the user to switch between different types of interest points. Functions are provided that can be used to test if it provides scale and/or orientation information on the interest point. The disadvantages of using this interface is that it prevents tight coupling between algorithms, leading to excessive computations. If descriptors are also being computed, consider using the DetectDescribePoint interface instead.

Example Code:

Concepts:

  • Point feature detection

Relevant Applets:

Related Examples:

Example Code

public class ExampleInterestPoint {

	public static <T extends ImageSingleBand>
	void detect( BufferedImage image , Class<T> imageType ) {
		T input = ConvertBufferedImage.convertFromSingle(image, null, imageType);

		// Create a Fast Hessian detector from the SURF paper.
		// Other detectors can be used in this example too.
		InterestPointDetector<T> detector = FactoryInterestPoint.fastHessian(
				new ConfigFastHessian(10, 2, 100, 2, 9, 3, 4));

		// find interest points in the image
		detector.detect(input);

		// Show the features
		displayResults(image, detector);
	}

	private static <T extends ImageSingleBand>
	void displayResults(BufferedImage image, InterestPointDetector<T> detector)
	{
		Graphics2D g2 = image.createGraphics();
		FancyInterestPointRender render = new FancyInterestPointRender();


		for( int i = 0; i < detector.getNumberOfFeatures(); i++ ) {
			Point2D_F64 pt = detector.getLocation(i);

			// note how it checks the capabilities of the detector
			if( detector.hasScale() ) {
				double scale = detector.getScale(i);
				int radius = (int)(scale* BoofDefaults.SCALE_SPACE_CANONICAL_RADIUS);
				render.addCircle((int)pt.x,(int)pt.y,radius);
			} else {
				render.addPoint((int) pt.x, (int) pt.y);
			}
		}
		// make the circle's thicker
		g2.setStroke(new BasicStroke(3));

		// just draw the features onto the input image
		render.draw(g2);
		ShowImages.showWindow(image, "Detected Features", true);
	}

	public static void main( String args[] ) {
		BufferedImage image = UtilImageIO.loadImage(UtilIO.pathExample("sunflowers.jpg"));
		detect(image, ImageFloat32.class);
	}
}