Example Non Maximum Suppression

From BoofCV
Revision as of 13:06, 17 January 2022 by Peter (talk | contribs)
Jump to navigationJump to search

Non-maximum suppression is a class of algorithm used to find local peaks and minimums inside a feature intensity image. This example demonstrations how to use efficient algorithms inside of BoofCV to quickly find extremes. Example Code:

Concepts:

  • Feature detection

Related Examples:

Example Code

/**
 * Non-maximum suppression is used to identify local maximums and/or minimums in an image feature intensity map. This
 * is a common step in feature detection. BoofCV includes an implementation of non-maximum suppression which is much
 * faster than the naive algorithm that is often used because of its ease of implementation. The following code
 * demonstrates how some of the tuning parameters affects the final output.
 *
 * @author Peter Abeles
 */
public class ExampleNonMaximumSupression {

	public static BufferedImage renderNonMax( GrayF32 intensity, int radius, float threshold ) {
		// Create and configure the feature detector
		NonMaxSuppression nonmax = FactoryFeatureExtractor.nonmax(new ConfigExtract(radius, threshold));

		// We will only searching for the maximums. Other variants will look for minimums or will exclude previous
		// candidate detections from being detected twice
		QueueCorner maximums = new QueueCorner();
		nonmax.process(intensity, null, null, null, maximums);

		// Visualize the intensity image
		BufferedImage output = new BufferedImage(intensity.width, intensity.height, BufferedImage.TYPE_INT_RGB);
		VisualizeImageData.colorizeSign(intensity, output, -1);

		// render each maximum with a circle
		Graphics2D g2 = output.createGraphics();
		g2.setColor(Color.blue);
		for (int i = 0; i < maximums.size(); i++) {
			Point2D_I16 c = maximums.get(i);
			VisualizeFeatures.drawCircle(g2, c.x, c.y, radius);
		}
		return output;
	}

	public static void main( String[] args ) {
		BufferedImage buffered = UtilImageIO.loadImageNotNull(UtilIO.pathExample("standard/boat.jpg"));

		GrayF32 input = ConvertBufferedImage.convertFrom(buffered, (GrayF32)null);

		// Compute the image gradient
		GrayF32 derivX = input.createSameShape();
		GrayF32 derivY = input.createSameShape();

		GImageDerivativeOps.gradient(DerivativeType.SOBEL, input, derivX, derivY, BorderType.EXTENDED);

		// From the gradient compute intensity of shi-tomasi features
		GeneralFeatureIntensity<GrayF32, GrayF32> featureIntensity =
				FactoryIntensityPoint.shiTomasi(3, false, GrayF32.class);

		featureIntensity.process(input, derivX, derivY, null, null, null);
		GrayF32 intensity = featureIntensity.getIntensity();

		ListDisplayPanel panel = new ListDisplayPanel();
		panel.addImage(buffered, "Input Image");
		// hack to just show intensity - no features can be detected
		panel.addImage(renderNonMax(intensity, 10, Float.MAX_VALUE), "Intensity Image");

		// Detect maximums with different settings and visualize the results
		panel.addImage(renderNonMax(intensity, 3, -Float.MAX_VALUE), "Radius 3");
		panel.addImage(renderNonMax(intensity, 3, 30000), "Radius 3  threshold");
		panel.addImage(renderNonMax(intensity, 20, -Float.MAX_VALUE), "Radius 10");
		panel.addImage(renderNonMax(intensity, 20, 30000), "Radius 10 threshold");

		ShowImages.showWindow(panel, "Non-Maximum Suppression", true);
	}
}